Provider setup index
Find pages by the provider you are connecting. A page may involve several providers, in which case it appears under each of them.
776 topic pages, distilled from 2266 real issuesHow this is counted
DeepSeek127
Codex auth.json overwritten, KEY mix-up, or official login lostUpgrade cc-switch to the latest version (v3.16.1 fixes the v3.16.0 KEY mix-up regression; v3.20.1/4.0 fix logout overwriting auth.json and losing… (Codex)settings.json configuration lost after switching providers (cc-switch)Extract shared fields into “Common Configuration” and enable writing them for each provider; v3.12.2 fixes the proxy-related BUG.Codex switching DeepSeek model error or only showing customEnable local routing and ensure Codex is logged into the official account to generate models_cache.json, restoring the model list.cc-switch Usage Statistics Differ from DeepSeek Official Site / Stop UpdatingUpgrade to v3.19.2 or later, then run “Rebuild Codex Usage” once on the “Usage Statistics” page to correct historical overcounts or missing records.Switching Codex providers overwrites config.tomlUpgrade cc-switch to v3.17.0+ and sync the MCP and other settings you want to keep into “General Configuration”; if an old snapshot still writes them back…schema must be a JSON Schema of 'typeThis is a conversion defect in older versions of CC Switch. Upgrade to v3.18.0 or later to fix it; no manual changes to config.toml are needed.Codex config.toml completely overwrittenMost related issues remain unresolved. Put custom sections in cc-switch’s common configuration or provider template, then upgrade to the latest version. (MCP)DeepSeek Claude Code Configuration and Switching IssuesFirst confirm the issue type: new terminals or sessions reload configuration. For abnormal DeepSeek cache hit rates, try v3.16.3; other variants remain…settings.json config lost after switching providersUse the "Shared Config Snippet" feature to extract and preserve custom fields like hooks/statusLine so they are carried over automatically on switch.Invalid schema for function — CodexUpgrade CC Switch to a version including the v3.18.0 fix for `parameters.type:null`; track the original issues for other `null` schema variants.The `content[].thinking` in thinking modeSome variants remain unresolved: upgrade to a version with the thinking history normalization fix; for Codex routed through DeepSeek, use a version with the…Codex model catalog template `gpt-5.5`Run `codex debug models --bundled > ~/.codex/models_cache.json` to generate the missing model cache, then switch the Codex route again."Failed to upload "image.png"" screenshot upload failedThis is not a cc-switch error. GitHub failed to upload the screenshot. Re-upload it or paste the error as text so maintainers can see it.truncation_policy.limit hardcoded to 10000 limits CodexIn cc-switch, manually set truncation_policy.limit in cc-switch-model-catalog.json to the model's actual context length as a temporary fix.api failed deserialize json body intoDepending on the error: roll Claude Code back, upgrade cc-switch, turn the local proxy off, or stop sending unsupported fields.Codex conversations with third-party models frequently stop automaticallycc-switch loses function call name/call_id during Responses API conversion. Connect DeepSeek directly to Codex, or upgrade cc-switch and retest.Provider switch not taking effect / Codex request failed (cc-switch)After switching providers such as DeepSeek in cc-switch, upgrade cc-switch and use its enable/save action as the sole way to write live configuration; do not…deepseek接入 模型勾选了1m: no 1M context option in Claude Code model pickerUpgrade cc-switch to v3.17.0+, leave the DeepSeek provider's ANTHROPIC_MODEL empty or check 1M there too, fill only Haiku/Sonnet/Opus mappings… (Claude Code)[SRV-006] connection errorUpgrade cc-switch to v3.16.4, which fixes proxy forwarding disconnects with a custom DeepSeek provider and billing falling back to the request model.Unexpected status 502 Bad Gateway：CodexSet `NO_PROXY=127.0.0.1,localhost,::1` in `~/.codex/.env` so Codex connects directly to the local CC Switch port.Provider API Key Overwritten: After Codex Restart or SwitchUpgrade to cc-switch v4.0.0 or later to prevent provider keys from overwriting each other and restore the official login state.Codex desktop cannot see modelsSign in to official ChatGPT, enable “Keep official login when switching to third parties,” and enable local routing. Restart Codex.Codex shows no model list and will not switch modelsTurn on keeping the official login when switching to a third party in cc-switch, and make sure you signed in to official ChatGPT in Codex once.Check failed Connection failedToggle or switch the system proxy, then fully restart cc-switch. If needed, upgrade macOS or update to v3.16.3+, preferably 4.x.Upstream HTTP 404 - DeepSeekResolve based on the scenario: if the endpoint path is the issue, complete the API address with the full API root (such as `.../v1`); if the model… (cc-switch)model does not support thinkingTurn the Proxy Rectifier off in cc-switch advanced settings, or upgrade to a fixed version. Qwen tool call loops and empty thinking blocks remain unresolved.auto mode cannot determine the safety: Claude Code errorUpgrade CC Switch first. If the error persists, check thinking parameters, model mapping, stop support, and the response format when stream is omitted.Model list is empty; Codex cannot select the DeepSeek modelUpgrade to CC Switch 4.0 and reapply the DeepSeek preset; model list display in the Codex desktop app also depends on its version support.Error from provider (Console Go): Upstream request failedCheck whether the model mapping was applied. If the target model is unavailable or the request body contains an unsupported tool_choice or image… (cc-switch)Claude Code /context Shows 200K Instead of 1MIn Claude Code, run /model claude-opus-4-8[1m], or set ANTHROPIC_DEFAULT_OPUS_MODEL to claude-opus-4-8[1m] in ~/.claude/settings.json.DeepSeek Responses conversion causes low cache hit rate, CodexFor a low cache hit rate with Codex and DeepSeek, check request logs and the proxy conversion path; no confirmed fix is available.Add CoderPlan provider preset for Claude CodeFeature request: add a CoderPlan provider preset to cc-switch. A Claude Code preset PR exists, but merge status is unconfirmed.Codex Desktop project list disappears or reasoning levels are reduced - CC Switch configuration issueMost related issues remain unresolved. After upgrading CC Switch or switching providers, changes to Codex’s model_provider field can make project history…tools.function.parameters.type is requiredUpgrade cc-switch to v3.18.0 to fix Codex tool parameter schemas rejected by strict upstreams with HTTP 400.Codex startup failure: missing supports_reasoning_summariesUpgrade cc-switch to v3.18.0 or later, then switch to the DeepSeek provider again to regenerate the model catalog. This automatically fills in the… (Codex)got 0 from 0 output items: Codex remote compaction failedUpgrade CC Switch to v4.0.0 or later. If you cannot upgrade yet, disable Remote Compaction and use Codex local compaction.Codex unavailable: reconnect failed 5 timesAfter switching providers, verify and re-enter the current provider's API Key; API Key overwrite reports continued in v3.16.3.DeepSeek usage statistics show the wrong cost / cache hit rateUpgrade to v3.16.3+, fixing input double-counting on the chat→message conversion path, streaming without usage, and wrong model billing attribution.502 Bad GatewayCodex/Claude Desktop reports 502 due to CLI proxy conflicts or app-server bypassing proxy. Remove proxy config or downgrade versions to fix.unknown variant `image_url`, expected `text`Do not send images to models that do not support image input. Use text only, or wait for/use the merged automatic retry mechanism that removes images.Invalid schema for functionCodex Desktop’s built-in automation_update tool schema lacks a top-level type, so DeepSeek rejects it. Upgrade CC Switch to a version that includes the fix.API Error: 400 Param Incorrect Claude CodeUpgrade CC Switch to v3.16.2+ to fix 400 errors caused by missing reasoning_content in MiMo thinking mode.missing field `call_id`: Codex automation HTTP 400No verified published fix is available. Normalize orphaned output without `call_id` during proxy forwarding and retain cross-task message content.Request body format invalid error - Claude CodeUpgrade cc-switch and enable the Rectifier, or add MAX_THINKING_TOKENS=0, for compatibility with third-party API formats.unexpected status 404 Not FoundUpgrade cc-switch to v3.16.4+ and set the upstream format explicitly to Chat Completions on the Codex provider form."deepseek-v4-flash" is not a modelMost related issues remain unresolved. Upgrade to CC Switch v4, enter the model's actual context window, and use [1m] only if it truly supports 1M.Failed to authenticate: Claude Desktop gateway token invalid — ClaudeSign out of the stale official account session in Claude Desktop and restart the app. If the error persists, switch to the default provider and then switch…Select Model Menu Still Shows Pre-Switch Model List - CodexClose existing Codex sessions, run `codex app-server daemon restart`, and upgrade CC Switch to v4.0.7 or later for detection and restart prompts.401 Unauthorized: Incorrect API key providedCodex via CC Switch to DeepSeek returns 401; manually add http_headers auth headers (Authorization: Bearer or x-api-key) in the config file.The model cannot discover any tools - CodexUpgrade cc-switch past 3.16.1 and edit cc-switch-model-catalog.json to set supports_search_tool to false for the DeepSeek models.codex never shows the ds model - CodexCodex desktop needs an official login state to show third-party models. Enable keep-official-login-when-switching in cc-switch and restart.cc-switch overrides DNS cache and resolvesDisable proxy takeover in the affected application to restore network connectivity; cc-switch does not modify system DNS records.It reports 401 in claude - Claude CodeTurn the cc-switch local proxy on to convert the API format, and make sure `ANTHROPIC_BASE_URL` is injected into the Claude Code process.The 'deepseek-v4-flash' modelEnable the keep-official-login option in cc-switch, confirm local route takeover is on, then restart Codex desktop.invalid tool_call_id (HTTP 400) — CodexIf tool calls fail, use a native Responses-compatible GPT model; v3.16.5 can be rolled back to v3.16.4.<tool_use_error>InputValidationErrorAn SSE parsing bug in the cc-switch conversion layer loses tool calls. Connect to the API directly, or switch to an official node, for now.Cache hit rate only 40% and extremely slow - Claude CodeDisable the dynamic attribution header in `.claude/settings.json`; if that fails, switch to a model with working caching and monitor the main issue.Not logged in · Please run /login ClaudeEdit `~/.claude.json` to move `PROXY_MANAGED` into the `approved` list, or answer yes at the prompt.The proxy takeover claude switch cannot be ticked in Claude Desktop modeIn claude-desktop mode, the takeover section hard-codes the Claude Code (claude) switch instead of the Claude Desktop one, so clicking does nothing.session isn't eligibleFor a non-Anthropic upstream or format conversion, set CLAUDE_CODE_AUTO_MODE_SERVER=0 in the provider env.Configuring DeepSeek in Codex errors about gpt-5.5Install the Codex CLI and run codex debug models --bundled to regenerate models_cache.json.Invalid schema: codex_app__automation_updateUpgrade CC Switch to v3.18.0 or newer to fix the 400 caused by type: null when Codex converts the DeepSeek protocol.The apply_patch tool is unavailable with the Codex proxy onUpgrade to cc-switch v3.16.2 or newer. On a relay, make sure the backend model matches what Codex expects or tool calls fail with routing on.X-OpenAI-Internal-Codex-Responses-LiteComment out the custom provider model_provider setting and return to the built-in openai provider; alternatively, temporarily switch to gpt-5.4.No such file or directory (os error 2)Change the model_catalog_json absolute path in config.toml to the relative cc-switch-model-catalog.json, or upgrade cc-switch.unexpected status 401 UnauthorizedUpgrade to CC Switch 4.0. On 3.x, declare a custom model_provider in config.toml so Codex does not ignore the proxy and connect directly to OpenAI.The `reasoning_content` in the thinking mode must be passed back to the APIConfigure DeepSeek as native mode instead of the default; if it still fails, set `model_reasoning_effort` to `none` in Codex.Gateway rejected model "claude-haiku-4-5"Enable routing based on the API type; for 火山方舟 Responses API, use the full URL and set `isFullUrl=true` to prevent the path from being appended again.Hermes cannot see any of these Skills (HERMES_HOME not recognized)CC Switch writes Symlinks to ~/.hermes/skills/; copy them to $HERMES_HOME/skills/ so Hermes sees them. The CN-version session issue has no user-side fix.HTTP 400 context length exceeded, CodexUpgrade to cc-switch 4.0. If the issue still recurs, provide the version and reproduction steps, then report it again.AbsolutePathBuf deserialized without a basemacOS Codex Desktop users can set model_catalog_json in config.toml to the absolute path of the model catalog file.Invalid pages parameter: "" Claude CodeSwitch the API format to anthropic, or upgrade cc-switch to fix the empty pages field from Responses conversion.Invalid schema for function 'web_search'Upgrade cc-switch to v3.17.0 or later. If upgrading is not possible, use the native DeepSeek Anthropic-compatible API.ANTHROPIC_DEFAULT_SONNET_MODEL not taking effectIn CC Switch local proxy mode, set the target model in ANTHROPIC_DEFAULT_SONNET_MODEL; the proxy maps this field, not *_MODEL_NAME.Unable to Call Tools Normally with Claude CodeModel responses vary on the NVIDIA channel; try deepseek-ai/deepseek-v4-flash, which was tested as stable, or switch providers.thinking options type cannot be disabledUpgrade cc-switch to v3.16.4 or later; the agent automatically strips the reasoning_effort parameter conflicting with DeepSeek.unexpected status 404 Not Found: Unknown error - CodexChange the DeepSeek official API request URL to one that includes `/v1`, then send the request through Codex again.Unexpected status 503 Service UnavailableWhen using v2ray, configure its routing to bypass LAN and mainland China domains, then access through CC Switch local routing again.Upstream request failed: Claude Code request failedFirst check the specific error returned by the upstream service. The provider must address regional unavailability;Volcengine Agent Plan Auto-Identify Model List Failed 404Manually modify config.toml to match official docs and add models in modelCatalog to bypass 404.Both ANTHROPIC_AUTH_TOKEN and ANTHROPIC_API_KEY setSwitching from DeepSeek or MiMo to Codex writes both authentication variables at once; this issue has no confirmed Codex fix.<thinking> tags output as body text in Claude CodeChanging the provider's apiFormat from openai_chat to anthropic, if supported upstream, can bypass inline <thinking> leakage.400 MissingParameter: input.typeNo verified fix is available. The reporter's direct tests returned 200 with "type": "message", but the proxy patch remains a proposal.Request exceeds the context windowNo verified solution was provided in the original post; this issue was closed as a duplicate related to catalog configuration and automatic compression.session isn't eligible because yourSet `CLAUDE_CODE_AUTO_MODE_SERVER` to `0` in the `env` of the CC Switch DeepSeek provider configuration, then confirm the key remains after switching providers.Cannot Read the Model List Correctly in Claude CodeThis issue concerns an older configuration that changed with cc-switch 4.0. If it still occurs after upgrading, report the version and reproduction steps.content[].thinking must be passed backTemporarily disable automatic updates and downgrade Claude Code to 2.1.150; the original post did not confirm a cc-switch fix.However, this session isn't eligibleUpdate to a version with the #7893 fix; it writes `CLAUDE_CODE_AUTO_MODE_SERVER` to settings.json when switching proxies./v1/models returns {"models":[]}A user restored 1M with a Python proxy returning model IDs with the [1m] suffix and context_window: 1048576; no reusable code or official fix was provided.Claude Desktop keeps timing outThe original post has no verified solution; users reported the settings.json suggestion in comments was ineffective.请求转发失败: 上游请求失败: client error (Connect)Temporarily bypass the CC Switch local proxy or switch networks to check whether DeepSeek is blocked. No final fix has been confirmed.Format conversion error: No choices in response - CC SwitchClinePass wraps the response in a data envelope, breaking parsing. Upgrade CC Switch to a version with the fix, or use a local proxy to unwrap response.data.Codex desktop app asks to log in again after restart - CodexUpgrade CC Switch from 3.16.0 to v3.16.2 or later to avoid overwriting Codex OAuth authentication when switching providers.After saving, the Codex config file still shows flashUpgrade to v3.17.0 or later and explicitly select deepseek-v4-pro in the “Default model” field.missing field `call_id`DeepSeek Responses rejected a Codex delegated task entry missing `call_id`; the issue did not confirm a user-side fix.thinking too short CodexStart a new Codex conversation to temporarily avoid it. The original post suggests filtering empty or overly short thinking; an official fix is unverified.missing field `call_id`Clean up orphaned heartbeat output entries when enabling requests through a proxy; this issue did not confirm that CC Switch has a fix.invalid thinkingKimi's deepseek-v4-flash rejects the current thinking parameter; the original post provides no verified solution.Usage Limit reached — Codex third-party provider quota exhaustedThis issue report says setting `requires_openai_auth` to `false` for a third-party provider may prevent Codex from checking official account quota and… (API)config.toml and auth.json not switching together — CodexUpgrade to CC Switch v3.16.1 or later; reports indicate the v3.16.1 pre-release resolved this issue.unknown type: tool_search_call — Codex session continuation returns 400Do not resume a thread containing tool_search_call history; start a new Codex thread before using the Responses upstream.ChatGPT cannot load config.toml: Codex DesktopRemove residual API login preferences from the Codex configuration and CC Switch saved configuration, then reopen the old conversation.Provider card and config.toml out of sync CodexMost related issues remain unresolved. Upgrade CC Switch from affected 3.20.4 and retry switching; if it persists, retain logs and configuration evidence.supported API model names… passed Old Model Name: CodexAfter switching providers, the old session still sends the original model name. Start a new Codex session, or restart the old session.Tool calls always hang: Codex function_call loses call_idA commenter confirmed non-empty id/name checks restore tool calls. The post also suggests bypassing CC Switch Responses translation to connect directly to…Messages with role 'tool' must be a response to a preceding message with 'tool_calls' (Codex)In cc-switch, toggle the "local routing" switch off and on again to temporarily recover; this is a proxy internal state bug that recurs.unsupported call: - CodexUpgrade cc-switch to v3.16.4+ to fix tool call name/id overwritten by empty deltas during streaming conversion.DeepSeek integration with cc-switch: Codex deep integration feature requestThis requests deep integration of DeepSeek with Codex. The original post has no verifiable solution; closure does not prove implementation.Failed to deserialize the JSON bodyDeepSeek rejects conversation history after an image has been sent. Start a new conversation to recover.Intermittent 502: Claude Code Using DeepSeekThe original post provides no verifiable solution. Provide the local network environment and related configuration to determine the cause.Test failed Claude CodeCheck the actual proxy port and set the CC Switch outbound proxy to that port; in this case, changing it to 7897 restored connectivity.Sessions may stop responding (JSONL stops being written) — Claude CodeUpdate CC Switch to the latest 4.0 release and retry. If it still hangs, temporarily bypass CC Switch and submit complete reproduction details.API key is empty: cc-switch DeepSeek balance queryUpgrade to v3.16.1 or later and verify again. The original issue did not confirm the fix.level "max" not supported, Codex DesktopSelect max for gpt-5.6-sol in Codex Desktop; ultra is not a separate reasoning effort.Local routing mode only changes the Fable model: Claude CodeUpgrade to 4.0 or later; local routing writes Haiku, Sonnet, Opus, and Fable as routing aliases.input_modalities contains only text: Codex cannot read imagesManually adding image to input_modalities reportedly enables image recognition, but switching models resets it. No verifiable lasting solution is available.Blank or missing model names in /models — OpenCode v2Remove empty name fields from the OpenCode configuration to restore names. Saving again in CC Switch brings the issue back; no official fix is documented.Invalid request parameters - CodexOpenCode-Go's mimo-v2.6-flash fails when it does not natively support `/responses`; this issue provides no verified fix.Only under the OpenCode agentUpdate to v3.16.4 or later to use OpenCode Go presets with API Key-only configuration in Claude Code, Claude Desktop, Codex, and OpenCode.Input should be a valid string: Codex image history 422Most related issues remain unresolved. Upgrade to cc-switch v3.20.4 and retry. If 422 persists, start a new session or roll back before the image.maximum context length is 1048565 tokens.Reduce PDF or message input. If the DeepSeek-compatible endpoint does not support PDF, stop uploading it, update CC Switch, and retry.v4.1 flash's input doesn't have ["image"]: PiUpgrade to v4.0.0 and add DeepSeek again from the preset, or manually set the model input to ["text", "image"].Pi provider list omits channels configured through /loginThe Pi /login channels report remains open. No verified solution is available; read-only display in CC Switch is only a proposal.CC Switch is currently a black box; users cannot customize processing logicThis is a feature request for programmable routing transformation scripts in CC Switch; the original post has no verifiable official solution.Requests keep failing after configuring a relay — Claude CodeEnable routing from the top bar, then apply it to Claude. “Enable full URL” does not enable routing. The reporter confirmed this fixed the issue.No tool output found for tool callNo official fix is confirmed. The reporter describes a local workaround that reorders tool calls and outputs. Follow the main issue for progress.Changed files cannot show diff below. CodexDo not automatically fetch 火山方舟 Coding Plan models. Manually use deepseek-v4-flash, then restart Codex and CC Switch.
GLM / Zhipu74
Codex switching DeepSeek model error or only showing customEnable local routing and ensure Codex is logged into the official account to generate models_cache.json, restoring the model list.cc-switch Usage Statistics Differ from DeepSeek Official Site / Stop UpdatingUpgrade to v3.19.2 or later, then run “Rebuild Codex Usage” once on the “Usage Statistics” page to correct historical overcounts or missing records.Codex config.toml completely overwrittenMost related issues remain unresolved. Put custom sections in cc-switch’s common configuration or provider template, then upgrade to the latest version. (MCP)DeepSeek Claude Code Configuration and Switching IssuesFirst confirm the issue type: new terminals or sessions reload configuration. For abnormal DeepSeek cache hit rates, try v3.16.3; other variants remain…The `content[].thinking` in thinking modeSome variants remain unresolved: upgrade to a version with the thinking history normalization fix; for Codex routed through DeepSeek, use a version with the…Upstream Chat Completions Streaming Response Error, Claude CodeCheck whether the issue is caused by local proxy conversion or the upstream endpoint; adjust the endpoint or disable the local proxy for the verified scenario.settings.json gets reset - Claude CodeWrite shared fields like plugins and hooks into "Common Config" and check write-back on switching; if takeover mode misbehaves, disable it, switch provider…api failed deserialize json body intoDepending on the error: roll Claude Code back, upgrade cc-switch, turn the local proxy off, or stop sending unsupported fields.deepseek接入 模型勾选了1m: no 1M context option in Claude Code model pickerUpgrade cc-switch to v3.17.0+, leave the DeepSeek provider's ANTHROPIC_MODEL empty or check 1M there too, fill only Haiku/Sonnet/Opus mappings… (Claude Code)Auth conflictLog out of the official Claude login, clear conflicting ANTHROPIC_* variables, set hasCompletedOnboarding to true in ~/.claude.json, and restart the terminal.Codex desktop cannot see modelsSign in to official ChatGPT, enable “Keep official login when switching to third parties,” and enable local routing. Restart Codex.model does not support thinkingTurn the Proxy Rectifier off in cc-switch advanced settings, or upgrade to a fixed version. Qwen tool call loops and empty thinking blocks remain unresolved.Invalid messages.content.type parameter; allowed value ['text']: CodexUpdate to v3.20.2 and re-import the Zhipu GLM preset; to send images in Codex, choose a vision-capable model.Error from provider (Console Go): Upstream request failedCheck whether the model mapping was applied. If the target model is unavailable or the request body contains an unsupported tool_choice or image… (cc-switch)Manually edited MCP configuration files are overwritten by cc-switchEdit MCP settings in cc-switch and upgrade to v3.13.0+; use “Common Configuration” to share settings.json plugins/hooks across providers.GLM 未找到可用模型 - Codex / CC SwitchUpgrade to cc-switch v3.20.4 and re-add the GLM provider from the preset with https://open.bigmodel.cn/api/v1 for Codex.unknown variant `image_url`, expected `text`Do not send images to models that do not support image input. Use text only, or wait for/use the merged automatic retry mechanism that removes images.invalid type: sequenceBack up and delete the old config.toml so Codex regenerates a default config, fixing the incompatible hooks format.The 'glm-5.3' model is not supported - CodexUpgrade to the latest version and disable proxy takeover. In 4.0.5 aggregation mode, start a new chat if old chats still fail randomly.Request body format invalid error - Claude CodeUpgrade cc-switch and enable the Rectifier, or add MAX_THINKING_TOKENS=0, for compatibility with third-party API formats."deepseek-v4-flash" is not a modelMost related issues remain unresolved. Upgrade to CC Switch v4, enter the model's actual context window, and use [1m] only if it truly supports 1M.built-in provider IDs: openai - CodexUpgrade CC Switch to v3.16.4+, delete the old config and save again. Enable unified Codex session history and use openai_base_url.The model cannot discover any tools - CodexUpgrade cc-switch past 3.16.1 and edit cc-switch-model-catalog.json to set supports_search_tool to false for the DeepSeek models.codex never shows the ds model - CodexCodex desktop needs an official login state to show third-party models. Enable keep-official-login-when-switching in cc-switch and restart.Calling GLM-5 reports error 1311 - cc-switchA provider permissions problem: enable GLM-5 / MiniMax / Kimi access with your provider, or ask them to confirm your subscription entitlements.ending with a leaked `</t` close-tagCaused by the cc-switch streaming converter not deduplicating and not filtering think tags. Update to a version containing the PR #6319 fix.Not found (404)Upgrade cc-switch to v3.16.3 for the 404 from /v1/ appended wrongly to a Base URL, and enable the media fallback so images cannot trip the breaker.exceeded retry limit, last status: 429 Too Many RequestsAn incorrect GLM preset URL can cause 429: delete the old GLM provider, create the official preset in cc-switch, confirm it uses a Coding Plan link, and check…Codex image requests return 400 through proxy: base64 inflation or invalid detailIf image history causes context overflow, upgrade to v3.19.0; if detail original is invalid, disable supports_image_detail_original for the affected model.invalid tool_call_id (HTTP 400) — CodexIf tool calls fail, use a native Responses-compatible GPT model; v3.16.5 can be rolled back to v3.16.4.401 UnauthorizedUpgrade to cc-switch 4.0.x, or clear the stale third-party OPENAI_API_KEY from ~/.codex/auth.json manually, restart Codex, and sign in again with ChatGPT OAuth.Bash failed due to the following issueSet the local GLM provider to OpenAI Chat format and enable routing so CC Switch converts and forwards Claude Code requests.ModelScope reports 401 token invalid - Claude CodeAdd ANTHROPIC_MODEL to the provider environment variables in cc-switch to resolve the 401. It is not synced to settings.json by design.[proxy::response_processor] [Claude] stream errorA stream decode error or timeout proxying Claude Code loses logs or records 0 tokens. Fix the model extraction fallback to show the right model.CC Switch online update freezes and becomes unresponsiveCC Switch online update can freeze under FlClash TUN mode. Switch to system proxy mode or manually download and reinstall.session isn't eligibleFor a non-Anthropic upstream or format conversion, set CLAUDE_CODE_AUTO_MODE_SERVER=0 in the provider env.API Error: 422 format conversion errorThe upstream NVIDIA NIM GLM-5.2 model intermittently returns an empty choices array. Turn on failover in cc-switch or retry the request.No Plans Displayed (Parsing Returns Empty) - Claude CodeUpgrade to cc-switch 4.0; the 3.19.1 usage template may no longer apply.HTTP 400 context length exceeded, CodexUpgrade to cc-switch 4.0. If the issue still recurs, provide the version and reproduction steps, then report it again.Invalid schema for function 'web_search'Upgrade cc-switch to v3.17.0 or later. If upgrading is not possible, use the native DeepSeek Anthropic-compatible API.Unable to Call Tools Normally with Claude CodeModel responses vary on the NVIDIA channel; try deepseek-ai/deepseek-v4-flash, which was tested as stable, or switch providers.Reasoning Level Not AdaptedRe-add the OpenCode Go preset after upgrading to get the correct reasoning_effort config.Volcengine Agent Plan Auto-Identify Model List Failed 404Manually modify config.toml to match official docs and add models in modelCatalog to bypass 404.Why is the value of ANTHROPIC_BASE_URL empty? Claude CodeThis is expected behavior: CC Switch does not inject variables into the shell; check the current configuration in `~/.claude/settings.json` directly.400 MissingParameter: input.typeNo verified fix is available. The reporter's direct tests returned 200 with "type": "message", but the proxy patch remains a proposal.The auto mode classifier is temporarily unavailableDisable the Anthropic Base URL proxy in CC Switch and restart Claude Code; one user in the issue reported that updating CLIProxyAPI resolved the problem.Invalid messages.content.type Parameter Claude CodeMost related issues remain unresolved. Upgrade to a version containing the #7656 fix so Chinese image-rejection errors trigger image fallback retries.Unable to apply reverse proxy to Claude Code VS Code pluginEnable “Apply to Claude Code plugin”; if the old provider is still active, check and remove the previous Zhipu relay settings in VS Code settings.json.However, this session isn't eligibleUpdate to a version with the #7893 fix; it writes `CLAUDE_CODE_AUTO_MODE_SERVER` to settings.json when switching proxies.Gateway /v1/models returned an unexpected body: Claude CodeNo verified solution is available; the local proxy must return an Anthropic `data` model-list format to Claude Code.Format conversion error: No choices in response - CC SwitchClinePass wraps the response in a data envelope, breaking parsing. Upgrade CC Switch to a version with the fix, or use a local proxy to unwrap response.data.Usage Limit reached — Codex third-party provider quota exhaustedThis issue report says setting `requires_openai_auth` to `false` for a third-party provider may prevent Codex from checking official account quota and… (API)Tool calls always hang: Codex function_call loses call_idA commenter confirmed non-empty id/name checks restore tool calls. The post also suggests bypassing CC Switch Responses translation to connect directly to…Failover Does Not Update API Base URL When Switching ProvidersUpgrade to 4.0 or the latest version and retry; failover forwards requests using the target provider's API URL and model mapping.GLM configuration overwritten during use in cc-switchIncrease streaming and non-streaming failover timeout thresholds, then upgrade to the version the maintainer said was updated.messages.content.type invalid parameter; allowed values ['text']Remove the image capability flag for glm-5.3 and use a text-only model catalog configuration.Model does not support image input CodexAdd an image-input modality declaration for GLM-5.3-Flash to avoid inheriting a model configuration containing only ["text"].“Bash tool is continuously searched for but cannot be loaded or invoked properly” - Claude CodeBypass CC Switch and configure GLM directly in Claude Code; or upgrade to CC Switch 4.0.13 and retest.Query Failed: cc-switch GLM (Zhipu) usage queryFor GLM, "Query Failed" may only mean usage query is unavailable; API forwarding can still work. Upgrade and retest, or use a custom template.GLM Team Plan usage query fails in Claude CodeUpgrade to v3.17.0 or later, then enter the organization ID and project ID in the 「Zhipu GLM Team」 template.Cannot disable glm-5.2 thinking - CodexSelecting glm-5.2 none thinking mode in Codex still enables thinking; this issue has no confirmed fix yet.Top-level model.provider not updated for Hermes DesktopThis report was closed as a duplicate. No verifiable solution was provided; track root-cause issue #5142.hooks configuration lost: cc-switch switches providersWrite hooks required by all providers into “General configuration”, and select it when editing configurations.Local routing mode only changes the Fable model: Claude CodeUpgrade to 4.0 or later; local routing writes Haiku, Sonnet, Opus, and Fable as routing aliases.This provider does not support fetching the model list CodexMake `models` entries compatible with `slug` and `id` so they do not affect valid `data` field parsing.max_tokens must be at least 1, got -1712The original post only suggested a lower-bound check for `max_tokens`; no verified fix or user-usable configuration is available.Blank or missing model names in /models — OpenCode v2Remove empty name fields from the OpenCode configuration to restore names. Saving again in CC Switch brings the issue back; no official fix is documented.Only under the OpenCode agentUpdate to v3.16.4 or later to use OpenCode Go presets with API Key-only configuration in Claude Code, Claude Desktop, Codex, and OpenCode.401 Unauthorized caused by PROXY_MANAGEDUpgrade to 4.0 or the latest version; use the real API key from the provider configuration for routing.model field is still overridden — Claude Code proxy takeoverUpdate to the latest CC Switch 4.0 release and retry. If it persists, temporarily remove the model fields from settings.json and report the result.content_block_start events fragmented hundreds of timesUpgrade to the next cc-switch version with b78192e8f; it ignores empty reasoning_content and merges consecutive text deltas.No tool output found for tool callNo official fix is confirmed. The reporter describes a local workaround that reorders tool calls and outputs. Follow the main issue for progress.image_detail is invalid: Codex StepFun Screenshot Request FailureUpgrade CC Switch to v3.20.4; it changes tool screenshot detail: "original" to the StepFun-accepted "auto".Unknown coding plan provider - Claude CodeUpgrade to v3.20.0 so the Agent Plan `/api/plan` endpoint is recognized correctly.
OpenAI (official)69
Codex auth.json overwritten, KEY mix-up, or official login lostUpgrade cc-switch to the latest version (v3.16.1 fixes the v3.16.0 KEY mix-up regression; v3.20.1/4.0 fix logout overwriting auth.json and losing… (Codex)Switching Codex providers overwrites config.tomlUpgrade cc-switch to v3.17.0+ and sync the MCP and other settings you want to keep into “General Configuration”; if an old snapshot still writes them back…Invalid schema for function — CodexUpgrade CC Switch to a version including the v3.18.0 fix for `parameters.type:null`; track the original issues for other `null` schema variants.Work browser access requires an authenticated accountLog in to the official account in the Codex/ChatGPT app with auth_mode=chatgpt, then use cc-switch routing mode (or keep official login in direct mode) to…Switching providers resets claude settings.json - Claude CodeBefore switching, extract the current provider’s configuration in “General Configuration” and select “Write to General Configuration” for the new provider.Codex provider writeback resets live UI settings from configUpgrade to cc-switch v4.0 or later. Provider switching updates only key fields and preserves the rest of ~/.codex/config.toml.Codex config.toml Overwritten by CC Switch, Plugins and Settings LostUpgrade to cc-switch v4.0.0; it removes full backup restoration and replaces only key fields when switching, preserving other Codex settings.Cannot use Google-email Codex account after opening cc switch - CodexA third-party key in ~/.codex/auth.json overrode ChatGPT login: add the official provider in CC Switch and switch back to restore login, and upgrade to…Provider API Key Overwritten: After Codex Restart or SwitchUpgrade to cc-switch v4.0.0 or later to prevent provider keys from overwriting each other and restore the official login state.Codex shows no model list and will not switch modelsTurn on keeping the official login when switching to a third party in cc-switch, and make sure you signed in to official ChatGPT in Codex once.Upstream HTTP 404 - DeepSeekResolve based on the scenario: if the endpoint path is the issue, complete the API address with the full API root (such as `.../v1`); if the model… (cc-switch)"Not logged in · Please run /login" - Claude CodeUpgrade cc-switch to v3.19.2+, or change ANTHROPIC_API_KEY to ANTHROPIC_AUTH_TOKEN in settings.json by hand to skip the approval prompt.auth.json not updated after switching channelCodex authentication mechanism refactor causes switching to fail. Check auth.json path and configuration per the official authentication retention guide.The 'gpt-6-sol' model is not supportedUpdate the Codex OAuth reported client version to one supported by the model and rebuild, or temporarily use gpt-5.6-sol / gpt-5.6-luna.HTTP 400 Codex /responses Request FailureCheck whether the provider supports Codex’s request format. If it natively supports the Responses API, set apiFormat to openai_responses to avoid conversion.Codex unavailable: reconnect failed 5 timesAfter switching providers, verify and re-enter the current provider's API Key; API Key overwrite reports continued in v3.16.3.invalid type: sequenceBack up and delete the old config.toml so Codex regenerates a default config, fixing the incompatible hooks format.The 'glm-5.3' model is not supported - CodexUpgrade to the latest version and disable proxy takeover. In 4.0.5 aggregation mode, start a new chat if old chats still fail randomly.Unknown parameterTurn the cc-switch unified session history switch off and move sessions tagged custom back to openai, fixing the namespace parameter error.Unsupported content type: redacted_thinkingFor OpenRouter routes, use a proxy to filter redacted_thinking. Codex/GPT routes have no confirmed fix; upgrade to the latest version and retest.ChatGPT authentication fails with global proxy enabled - CodexUpgrade cc-switch to v3.16.4+ and enter a working proxy address under 「设置 → 路由 → 全局出站代理」, then retry Codex/ChatGPT authentication.built-in provider IDs: openai - CodexUpgrade CC Switch to v3.16.4+, delete the old config and save again. Enable unified Codex session history and use openai_base_url.unexpected status 404 Not Found — CodexChange `requires_openai_auth` for the custom provider in `config.toml` to `false`, then start a new session to avoid sending `gpt-reserve`.401 Unauthorized: Incorrect API key providedCodex via CC Switch to DeepSeek returns 401; manually add http_headers auth headers (Authorization: Bearer or x-api-key) in the config file.codex never shows the ds model - CodexCodex desktop needs an official login state to show third-party models. Enable keep-official-login-when-switching in cc-switch and restart.Tray menu does not show current Plan allowance when multiple ChatGPT accounts are linked - CodexUpgrade cc-switch: the tray menu failing to show allowance with multiple linked ChatGPT accounts is a confirmed bug, fixed in commit e53ab21 and released…The 'deepseek-v4-flash' modelEnable the keep-official-login option in cc-switch, confirm local route takeover is on, then restart Codex desktop.Error running remote compact task error - CodexUpgrade cc-switch to v3.12.0 or later. It adds forwarding support for the Codex `/responses/compact` route in the local proxy and can resolve 404 or 502 errors.Gtk-CRITICAL: gtk_widget_get_scale_factor failedOn a GTK assertion failure running cc-switch under Linux or WSL2, edit settings.json to turn silent startup off, or clean the leftovers and reinstall.invalid_authentication_error Test Failure HTTP 401An unreplaced {{apiKey}} variable in the usage query script causes 401. Upgrade cc-switch to the latest version; temporarily enter the Key as a literal.Not logged in ? Please run /loginEnable local proxy takeover for Claude Code, point ANTHROPIC_BASE_URL to http://127.0.0.1:15721, and set proxy-managed authentication placeholders.<tool_use_error>InputValidationErrorAn SSE parsing bug in the cc-switch conversion layer loses tool calls. Connect to the API directly, or switch to an official node, for now.401 UnauthorizedUpgrade to cc-switch 4.0.x, or clear the stale third-party OPENAI_API_KEY from ~/.codex/auth.json manually, restart Codex, and sign in again with ChatGPT OAuth.HTTP 401 Unauthorized - CC Switch AnthropicUpgrade cc-switch to a fixed build so it sends the x-api-key and anthropic-version headers Anthropic requires.url is not supported for stdio, CodexUpgrade to v4.0.0 or later, then resync the servers on the MCP page to clear the old type field.[mcp_servers.node_repl] written back automatically by CC SwitchTurn off or delete node_repl in MCP management at the top right of CC Switch to stop it being written back into the Codex config file.unknown configuration fieldRemove the `type` field from the Codex MCP configuration; an unfixed cc-switch will write it back during the next sync.Session expired - Codex OAuth multiple accountsComplete Codex OAuth authorization again in a separate browser session signed in only to the target account, then restart CC Switch to verify.Still requires login to chatgpt: CodexRetest after upgrading to CC Switch 4.0; third-party provider configuration should be written to Codex's config.toml.Session expired: Codex OAuth session expiredUpgrade to cc-switch v4.0.0 or later to prevent old snapshots from overwriting new login credentials.Wrong model called / stop: cancel task, CodexAfter upgrading to 4.0 or the latest version, try again. Version 4.0 rewrote how Codex configuration is written to prevent the wrong model from being called…unexpected status 401 UnauthorizedUpgrade to CC Switch 4.0. On 3.x, declare a custom model_provider in config.toml so Codex does not ignore the proxy and connect directly to OpenAI.expected value at line 1 column 1Upgrade to cc-switch v3.14.1 or newer to fix the JSON parse failure caused by Codex OAuth forcing an SSE stream.AbsolutePathBuf deserialized without a basemacOS Codex Desktop users can set model_catalog_json in config.toml to the absolute path of the model catalog file.Reasoning effort dropped, Claude CodeWait for the next release containing the merged Responses proxy fix so reasoning.effort can be forwarded after model mapping.Gateway /v1/models returned an unexpected body: Claude CodeNo verified solution is available; the local proxy must return an Anthropic `data` model-list format to Claude Code.WebSearch returns 0 results: Claude CodeTemporarily disable built-in WebSearch. Use Brave Search MCP to find pages, then use WebFetch to read them.Official account not displaying normally in the codex lower-left corner - CodexThe original post provides no verifiable solution; follow the main issue #3566 referenced by the maintainer.After switching, opening codex still uses the old account — CodexAfter switching OAuth accounts, fully restart the Codex background service. For CLI, run `codex app-server daemon stop`, then reopen Codex.auth.json Is Force-Replaced, Making Codex UnusableUpgrade to a later version that includes the fix. Before upgrading, disable Codex takeover, sign in to official ChatGPT, then re-enable takeover.Codex Provider switching fails: unable to resolve account IDCopying codex_oauth_auth.json from device A to device B restored switching for the user. This is a temporary workaround; no official fix is recorded.OpenAI OAuth account/quota UI disappearsUpgrade to CC Switch 4.0 and confirm the official OAuth login is still stored on disk; local routing can still forward requests through a custom endpoint.404 Not Found: /v1/live, Codex Desktop VoiceDisable the Codex proxy takeover in CC Switch, restore the official Codex configuration, and fully restart the desktop app.Usage Limit reached — Codex third-party provider quota exhaustedThis issue report says setting `requires_openai_auth` to `false` for a third-party provider may prevent Codex from checking official account quota and… (API)Configured models not visible - CodexSign in to Codex with an official OpenAI account, retain the official login when switching providers, resave, and restart. Some API Key users report no fix.Codex custom endpoint silently redirected to chatgpt.com (withdrawn)Report withdrawn: the redirect affected only an unmerged commit. Routing works in existing versions; no configuration changes are needed based on this report.Unknown parameterUpdate to the replay allowlist fix so Claude Code filters the reasoning-envelope status field when resuming Codex OAuth sessions.exceeded the 64000/128000 output tokenRemove or rewrite the rejected MCP tool JSON Schema `pattern` to prevent Codex OAuth from falsely reporting an output token limit.ChatGPT cannot load config.toml: Codex DesktopRemove residual API login preferences from the Codex configuration and CC Switch saved configuration, then reopen the old conversation.{"models":[]}: Codex OpenAI official proxy model catalog is emptyTemporarily place the Codex-format catalog file inside CODEX_HOME and reference it in config.toml; the official proxy forwarding fix is unconfirmed.Buttons All Grayed Out and Cannot Send: Codex DesktopNo confirmed fix. Verify the request path and protocol, enable local routing, and test a new session; official quota status may still limit Desktop.Codex still loads skills from ~/.agents/skills after being disabledCheck the skill storage location. If a copy is in ~/.agents/skills, Codex will keep loading it; disable or remove that copy yourself.Requests show zero duration/no first-outputThis issue remains open, with no confirmed user-side fix; missing request timing should remain unknown rather than display as 0.experimental_bearer_token = "PROXY_MANAGED"Upgrade to CC-Switch v3.16.2 or later. If it persists, sign in to Codex OAuth again and reopen the report as requested.Request timeout (504/502) - Codex local routing directly to OpenAIFirst resolve the local routing timeout to chatgpt.com; this issue provides no confirmed additional configuration fix.Only the Last Logged-in Codex Account Is RetainedAuthentication for multiple official accounts in the same ChatGPT Team overwrites each other. No verifiable fix was provided; follow main issue #2245.maximum context length is 1048565 tokens.Reduce PDF or message input. If the DeepSeek-compatible endpoint does not support PDF, stop uploading it, update CC Switch, and retry.[DISCUSS] Support for pi coding agentThis is a feature request merged into tracking issue #1855; the original post provides no verifiable solution.Optional isolation forced output Claude CodeSet `strict: false` on ordinary function tools during Responses conversion and preserve the existing required fields in `required`.
Gemini / Google62
settings.json configuration lost after switching providers (cc-switch)Extract shared fields into “Common Configuration” and enable writing them for each provider; v3.12.2 fixes the proxy-related BUG.cc-switch Usage Statistics Differ from DeepSeek Official Site / Stop UpdatingUpgrade to v3.19.2 or later, then run “Rebuild Codex Usage” once on the “Usage Statistics” page to correct historical overcounts or missing records.Switching providers resets claude settings.json - Claude CodeBefore switching, extract the current provider’s configuration in “General Configuration” and select “Write to General Configuration” for the new provider.EGL_BAD_PARAMETER: CC Switch White Screen on LinuxFor affected Linux AppImage environments, preload the host libwayland-client first. If the white screen remains, try the .deb package.skills installed to the wrong path - cc-switchOlder cc-switch installs skills into ~/.claude/skills/skills, which Claude Code ignores. Upgrade, or move the contents up to ~/.claude/skills/.Codex provider writeback resets live UI settings from configUpgrade to cc-switch v4.0 or later. Provider switching updates only key fields and preserves the rest of ~/.codex/config.toml.Cannot use Google-email Codex account after opening cc switch - CodexA third-party key in ~/.codex/auth.json overrode ChatGPT login: add the official provider in CC Switch and switch back to restore login, and upgrade to…Function call is missingUpgrade cc-switch. The Gemini API requires thought_signature in tool calls, which older versions did not add when forwarding requests.duplicate entry with key "mcp_servers" Hermes configuration errorUpgrade to cc-switch v3.16.3 or later, then switch providers to automatically clean up historical duplicate keys.MiniMax usage not showing; test script reports “no result returned” (cc-switch)MiniMax changed its balance endpoint, breaking the old parser. Upgrade to cc-switch v3.16.2 or later; older versions can temporarily use a custom query…Unknown name "prompt_cache_key" - cc-switchUpgrade cc-switch to v3.16.2+ to fix prompt_cache_key injection, or use an OpenRouter endpoint that natively speaks the Claude Code format.Kaspersky virus detection deletes ccswitch — Gemini CLIAdd CC Switch to the security allow list in Windows security software, then use Gemini CLI.Gemini CLI session history will not display - Gemini CLIGemini CLI 0.39.1 changed the history path. With the consumer service being retired, the old path will not be fixed — wait for Antigravity CLI support.Restarting the terminal and running gemini asks for an api-key - Gemini CLIUpdate cc-switch for the save-clears-config bug. If it still fails, that is a Gemini CLI read bug — put the environment variables in bash_profile.Manually edited MCP configuration files are overwritten by cc-switchEdit MCP settings in cc-switch and upgrade to v3.13.0+; use “Common Configuration” to share settings.json plugins/hooks across providers.Not installed / not installed or not executable - CC SwitchUpgrade CC Switch to v3.20.0+ for custom npm prefix and macOS detection failures; the false "several installations" on Windows needs v3.16.5+.not installed or not executableUpgrade CC Switch to v3.16.2+, or set the CLI install directory by hand in settings, and clear noise printed by your shell at startup.A default Gemini config appears and cannot be deleted - CC Switchdefault is the expected provider CC Switch creates by importing your existing live config. Switch the current provider in settings first, then delete it.JSON parse error: expected value at line 1Upgrade cc-switch to fix JSON corruption from concurrent writes, and add custom fields to general config so a switch cannot overwrite them.selected model (kimi-k2.5) may not exist errorRemove the trailing /v1 from the API address and use the Anthropic-compatible interface. If an old session contains images, run /compact or switch models.Claude code does not support gemini - CC Switchcc-switch cannot currently use Gemini as a Claude Code provider — wait for the local proxy feature, or use a third-party proxy project.The model cannot discover any tools - CodexUpgrade cc-switch past 3.16.1 and edit cc-switch-model-catalog.json to set supports_search_tool to false for the DeepSeek models.Function call is missingUse the community cc-switch-gemini-fixer patch, or inject skip_thought_signature_validator into history messages, to bypass the signature validation.Gtk-CRITICAL: gtk_widget_get_scale_factor failedOn a GTK assertion failure running cc-switch under Linux or WSL2, edit settings.json to turn silent startup off, or clean the leftovers and reinstall.Symlink creation failed: cc-switch skill sync error or target app cannot see skillOn Windows, symlink creation requires administrator privileges or Developer Mode; run cc-switch as administrator. Hermes/Pi have separate path and scanning…Not logged in ? Please run /loginEnable local proxy takeover for Claude Code, point ANTHROPIC_BASE_URL to http://127.0.0.1:15721, and set proxy-managed authentication placeholders.Local Claude CLI installed but Settings says it isn'tUpgrade CC Switch to v3.16.5 or later to fix false installation conflicts and garbled paths on Windows.HTTP 401 Unauthorized - CC Switch AnthropicUpgrade cc-switch to a fixed build so it sends the x-api-key and anthropic-version headers Anthropic requires.Codex CLI failure: CC SwitchUpgrade cc-switch to v3.16.3 or later to fix blockers in local environment detection and the upgrade flow.CC Switch online update freezes and becomes unresponsiveCC Switch online update can freeze under FlClash TUN mode. Switch to system proxy mode or manually download and reinstall.The proxy takeover claude switch cannot be ticked in Claude Desktop modeIn claude-desktop mode, the takeover section hard-codes the Claude Code (claude) switch instead of the Claude Desktop one, so clicking does nothing.session isn't eligibleFor a non-Anthropic upstream or format conversion, set CLAUDE_CODE_AUTO_MODE_SERVER=0 in the provider env.Unable to resolve apiKey and baseUrl variablesUse `{{ apiKey }}` or `{{baseUrl}}` in the custom usage query template and verify that `isValid: false` is treated as a failure.Function call is missing a thought_signatureUpgrade to the latest CC Switch 4.0 release and retry. If it persists, report the version, system, full reproduction steps, and logs.The Gemini CLI reading the project .env breaks cc-switchWrap the gemini command in PowerShell to load the ~/.gemini/.env environment variables by hand.cc-switch keeps saying there is an environment variable conflict and then deletes the variable Gemini CLIIgnore the environment variable conflict prompt to keep your system variables, or upgrade cc-switch to v3.12.3 to fix saving in Official mode.[GEMINI-SYNC] Never Appears in the Log: Gemini CLIUpgrade to cc-switch v4.0.1; this version fixes Gemini CLI `.jsonl` session file detection and incremental record parsing.Kimi balance query unavailableUpdate cc-switch to the latest version; the backend has automatically fixed Kimi balance recognition.CC Switch flashes once when clicking any buttonOn Windows, use WebView2 152.0.4191.62 Fixed Version only for CC Switch to avoid click flashing with WebView2 153.0.4234.32.http://127.0.0.1:15721 proxy configuration stuckUpgrade to CC Switch v3.16.2 to fix proxy hangs from corrupted backups, or use "Replace Config" mode to manually set env vars.Upstream request failed: Claude Code request failedFirst check the specific error returned by the upstream service. The provider must address regional unavailability;Main Window Never Appears; CC Switch Fails to Start on WindowsCheck Windows updates and WebView2 arguments first; the issues do not confirm a general fix for all cases.API Error: Connection refusedIn Windows WSL2 NAT mode, use socat to forward WSL's 127.0.0.1:15721 to the host CC Switch routing port.CC-Switch Cannot Recognize Antigravity Sessions, Token Statistics Are 0This is an unresolved feature request; no merged and verifiable solution is available yet.No new usage record is createdMost related issues remain unresolved. Upgrade to a cc-switch version containing the #7740 fix for Gemini streaming usage parsing and model attribution.Crash on Launch: CC Switch Windows Startup CrashReproduce on the latest version and provide the CC Switch, Windows, and WebView2 versions, along with the latest logs and Event Viewer error details.Codex settings become the Claude valuesCodex failover settings are overwritten by Claude settings after a restart. No verified fix is available in the original report; follow the source issue.Provider card and config.toml out of sync CodexMost related issues remain unresolved. Upgrade CC Switch from affected 3.20.4 and retry switching; if it persists, retain logs and configuration evidence.Prompt files are never rebuiltPrompt, live configuration, and caches are not fully rebuilt after database restoration. No verifiable fix is available; follow the merged discussion.API key is empty: cc-switch DeepSeek balance queryUpgrade to v3.16.1 or later and verify again. The original issue did not confirm the fix.Empty display, no sessions. Gemini CLIUpgrade to a CC Switch version that includes .jsonl session-reading support to display sessions from the newer Gemini CLI.Native Grok Build Usage Import Missing - CC Switch UsageMost related issues remain unresolved. Upgrade to v3.19.0 for automatic Grok Build usage import in official OAuth mode;CORS MisconfigurationNo verifiable fix is provided. Restricting CORS, removing the CORS layer, and adding a local token are reporter suggestions; no fixed version is evidenced.mcode session usage import: model stays unknown, latency stays 0, and pricing failsNo confirmed user-side fix is available; update the importer to backfill model and latency from telemetry or session data.Text-only: MiMo-V2.5-Pro in Gemini CLIMiMo-V2.5-Pro is text-only. Use mimo-v2.5 instead, or a V2.6 model that supports images.Only under the OpenCode agentUpdate to v3.16.4 or later to use OpenCode Go presets with API Key-only configuration in Claude Code, Claude Desktop, Codex, and OpenCode.[DISCUSS] Support for pi coding agentThis is a feature request merged into tracking issue #1855; the original post provides no verifiable solution.CC Switch is currently a black box; users cannot customize processing logicThis is a feature request for programmable routing transformation scripts in CC Switch; the original post has no verifiable official solution.Cache hit rate for the current request of an individual queryUpgrade to v4.0.14, then select the request log “Hit Rate” column under “Settings” on the Usage page.Cannot start normally: window unresponsive, system requires force quitUpdate CC Switch to 4.0 or the latest version and retry. If it persists, provide the full version, steps, logs, and screenshots.White flash when switching tabs, CC SwitchThe issue may be related to a WebView2 153.x update. If it continues after upgrading to CC Switch v4.0, provide the version and reproduction steps.cmd.exe console window flashesUpdate to 4.0 or the latest version and retry. If it persists, submit the version, system details, reproduction steps, and logs or screenshots.
Kimi / Moonshot52
settings.json configuration lost after switching providers (cc-switch)Extract shared fields into “Common Configuration” and enable writing them for each provider; v3.12.2 fixes the proxy-related BUG.Switching Codex providers overwrites config.tomlUpgrade cc-switch to v3.17.0+ and sync the MCP and other settings you want to keep into “General Configuration”; if an old snapshot still writes them back…Codex config.toml completely overwrittenMost related issues remain unresolved. Put custom sections in cc-switch’s common configuration or provider template, then upgrade to the latest version. (MCP)settings.json config lost after switching providersUse the "Shared Config Snippet" feature to extract and preserve custom fields like hooks/statusLine so they are carried over automatically on switch.Invalid schema for function — CodexUpgrade CC Switch to a version including the v3.18.0 fix for `parameters.type:null`; track the original issues for other `null` schema variants.The `content[].thinking` in thinking modeSome variants remain unresolved: upgrade to a version with the thinking history normalization fix; for Codex routed through DeepSeek, use a version with the…"Failed to upload "image.png"" screenshot upload failedThis is not a cc-switch error. GitHub failed to upload the screenshot. Re-upload it or paste the error as text so maintainers can see it.Codex Kimi HTTP 400 moonshot flavored json schemaUpgrade CC Switch to the latest version; the proxy automatically cleans Kimi-incompatible `$ref` sibling keywords and tool schemas missing root-level `type`.401 Invalid Authentication - Claude CodeCheck the API key and authentication variables required by your provider, and confirm Claude Code's actual authentication field contains the correct key.Codex conversations with third-party models frequently stop automaticallycc-switch loses function call name/call_id during Responses API conversion. Connect DeepSeek directly to Codex, or upgrade cc-switch and retest.Provider switch not taking effect / Codex request failed (cc-switch)After switching providers such as DeepSeek in cc-switch, upgrade cc-switch and use its enable/save action as the sole way to write live configuration; do not…Codex desktop cannot see modelsSign in to official ChatGPT, enable “Keep official login when switching to third parties,” and enable local routing. Restart Codex.Upstream HTTP 404 - DeepSeekResolve based on the scenario: if the endpoint path is the issue, complete the API address with the full API root (such as `.../v1`); if the model… (cc-switch)Add CoderPlan provider preset for Claude CodeFeature request: add a CoderPlan provider preset to cc-switch. A Claude Code preset PR exists, but merge status is unconfirmed.Codex Desktop project list disappears or reasoning levels are reduced - CC Switch configuration issueMost related issues remain unresolved. After upgrading CC Switch or switching providers, changes to Codex’s model_provider field can make project history…tools.function.parameters.type is requiredUpgrade cc-switch to v3.18.0 to fix Codex tool parameter schemas rejected by strict upstreams with HTTP 400.got 0 from 0 output items: Codex remote compaction failedUpgrade CC Switch to v4.0.0 or later. If you cannot upgrade yet, disable Remote Compaction and use Codex local compaction.kimi k2 reports 404 - CC SwitchA 404 from the connection test does not mean Kimi is unavailable. Set ANTHROPIC_BASE_URL to https://api.kimi.com/coding/ and verify by running Claude Code.API Error: 400 Param Incorrect Claude CodeUpgrade CC Switch to v3.16.2+ to fix 400 errors caused by missing reasoning_content in MiMo thinking mode.tools.function.parameters is not a validUpgrade to a later version containing #6863; it converts `$ref` with sibling keywords into Moonshot-compatible `allOf`.missing field `call_id`: Codex automation HTTP 400No verified published fix is available. Normalize orphaned output without `call_id` during proxy forwarding and retain cross-task message content.selected model (kimi-k2.5) may not exist errorRemove the trailing /v1 from the API address and use the Anthropic-compatible interface. If an old session contains images, run /compact or switch models."deepseek-v4-flash" is not a modelMost related issues remain unresolved. Upgrade to CC Switch v4, enter the model's actual context window, and use [1m] only if it truly supports 1M.Failed to authenticate: Claude Desktop gateway token invalid — ClaudeSign out of the stale official account session in Claude Desktop and restart the app. If the error persists, switch to the default provider and then switch…Calling GLM-5 reports error 1311 - cc-switchA provider permissions problem: enable GLM-5 / MiniMax / Kimi access with your provider, or ask them to confirm your subscription entitlements.ending with a leaked `</t` close-tagCaused by the cc-switch streaming converter not deduplicating and not filtering think tags. Update to a version containing the PR #6319 fix.invalid_authentication_error Test Failure HTTP 401An unreplaced {{apiKey}} variable in the usage query script causes 401. Upgrade cc-switch to the latest version; temporarily enter the Key as a literal.The user sent an empty message (Claude Code connected to Kimi — repeated empty messages)No official fix: /export the session and continue in a new one; bypass cc-switch so Claude Code connects directly to Kimi to find which layer sends empty…Unsupported parameter(s): prompt_cache_keyUpgrade cc-switch to the latest version, fixing prompt_cache_key not being stripped automatically when forwarding to the Nvidia API.session isn't eligibleFor a non-Anthropic upstream or format conversion, set CLAUDE_CODE_AUTO_MODE_SERVER=0 in the provider env.unexpected status 401 UnauthorizedIn Codex config.toml, set the AUTO COMPACT threshold below Kimi's 256K limit, verify model mapping, and reduce unnecessary MCP usage.Upstream HTTP 404 cc-switchUpgrade cc-switch to remove ?beta=true from NVIDIA requests. Kimi does not support Codex's Responses API; use Claude Code or a local adapter.Invalid schema for function 'web_search'Upgrade cc-switch to v3.17.0 or later. If upgrading is not possible, use the native DeepSeek Anthropic-compatible API.invalid_request_error Kimi For CodingSet ENABLE_TOOL_SEARCH=false in the Kimi configuration in cc-switch.Kimi balance query unavailableUpdate cc-switch to the latest version; the backend has automatically fixed Kimi balance recognition.Nvidia check failed: Not found (404)Upgrade cc-switch to v3.16.3+; new version uses lightweight HTTP probing, fixing NVIDIA interface false 404 errors.thinking enabled but reasoning_content missingDisable Thinking mode in Claude Code, or upgrade cc-switch to fix Kimi/Moonshot tool call errors.Volcengine Agent Plan Auto-Identify Model List Failed 404Manually modify config.toml to match official docs and add models in modelCatalog to bypass 404.Bash(...) Waiting... Claude CodeDowngrade Claude Code to v2.1.104 or v2.0.64 (admin required), or switch to a compatible provider such as Alibaba Cloud Bailian.However, this session isn't eligibleUpdate to a version with the #7893 fix; it writes `CLAUDE_CODE_AUTO_MODE_SERVER` to settings.json when switching proxies.Format conversion error: No choices in response - CC SwitchClinePass wraps the response in a data envelope, breaking parsing. Upgrade CC Switch to a version with the fix, or use a local proxy to unwrap response.data.Requests Keep Sending Continuously - Codex appFirst check proxy logs to confirm whether the ongoing requests are coming from Codex app; the original post does not provide a verifiable solution.invalid thinkingKimi's deepseek-v4-flash rejects the current thinking parameter; the original post provides no verified solution.Usage Limit reached — Codex third-party provider quota exhaustedThis issue report says setting `requires_openai_auth` to `false` for a third-party provider may prevent Codex from checking official account quota and… (API)supported API model names… passed Old Model Name: CodexAfter switching providers, the old session still sends the original model name. Start a new Codex session, or restart the old session.unsupported call: - CodexUpgrade cc-switch to v3.16.4+ to fix tool call name/id overwritten by empty deltas during streaming conversion.Sessions may stop responding (JSONL stops being written) — Claude CodeUpdate CC Switch to the latest 4.0 release and retry. If it still hangs, temporarily bypass CC Switch and submit complete reproduction details.Blank or missing model names in /models — OpenCode v2Remove empty name fields from the OpenCode configuration to restore names. Saving again in CC Switch brings the issue back; no official fix is documented.Only under the OpenCode agentUpdate to v3.16.4 or later to use OpenCode Go presets with API Key-only configuration in Claude Code, Claude Desktop, Codex, and OpenCode.Pi provider list omits channels configured through /loginThe Pi /login channels report remains open. No verified solution is available; read-only display in CC Switch is only a proposal.CC Switch is currently a black box; users cannot customize processing logicThis is a feature request for programmable routing transformation scripts in CC Switch; the original post has no verifiable official solution.No tool output found for tool callNo official fix is confirmed. The reporter describes a local workaround that reorders tool calls and outputs. Follow the main issue for progress.
Third-party relay42
config.toml content was reset in Codex (cc-switch)Upgrade to v4.0.0+: older versions overwrote the entire config.toml, losing MCP and projects settings. Newer versions replace only key fields and preserve the…Codex model catalog template `gpt-5.5`Run `codex debug models --bundled > ~/.codex/models_cache.json` to generate the missing model cache, then switch the Codex route again."Failed to upload "image.png"" screenshot upload failedThis is not a cc-switch error. GitHub failed to upload the screenshot. Re-upload it or paste the error as text so maintainers can see it.Cannot use Google-email Codex account after opening cc switch - CodexA third-party key in ~/.codex/auth.json overrode ChatGPT login: add the official provider in CC Switch and switch back to restore login, and upgrade to…Provider switch not taking effect / Codex request failed (cc-switch)After switching providers such as DeepSeek in cc-switch, upgrade cc-switch and use its enable/save action as the sole way to write live configuration; do not…deepseek接入 模型勾选了1m: no 1M context option in Claude Code model pickerUpgrade cc-switch to v3.17.0+, leave the DeepSeek provider's ANTHROPIC_MODEL empty or check 1M there too, fill only Haiku/Sonnet/Opus mappings… (Claude Code)Function call is missingUpgrade cc-switch. The Gemini API requires thought_signature in tool calls, which older versions did not add when forwarding requests.missing information - cc-switchA roundup of documentation gaps. Each case needs its own clarification, or waiting for the official documentation to fill in.Provider API Key Overwritten: After Codex Restart or SwitchUpgrade to cc-switch v4.0.0 or later to prevent provider keys from overwriting each other and restore the official login state.Codex shows no model list and will not switch modelsTurn on keeping the official login when switching to a third party in cc-switch, and make sure you signed in to official ChatGPT in Codex once.Check failed Connection failedToggle or switch the system proxy, then fully restart cc-switch. If needed, upgrade macOS or update to v3.16.3+, preferably 4.x.model does not support thinkingTurn the Proxy Rectifier off in cc-switch advanced settings, or upgrade to a fixed version. Qwen tool call loops and empty thinking blocks remain unresolved.Claude Code /context Shows 200K Instead of 1MIn Claude Code, run /model claude-opus-4-8[1m], or set ANTHROPIC_DEFAULT_OPUS_MODEL to claude-opus-4-8[1m] in ~/.claude/settings.json.Relay Link Speed Test Fails Without VPN - CC SwitchNo single confirmed fix is available. Provide a reproducible configuration, request target, and startup logs.Claude Official check failed: could not extract base_url - CodexCC Switch proxying and load balancing do not support an official Claude subscription — use a third-party API key. Forcing it errors and risks a ban.Codex Desktop project list disappears or reasoning levels are reduced - CC Switch configuration issueMost related issues remain unresolved. After upgrading CC Switch or switching providers, changes to Codex’s model_provider field can make project history…Usage query requires same-origin requests in cc-switchIn cc-switch, switch to the generic or custom template and enter the usage query URL directly in the request address to bypass the same-origin restriction.ChatGPT authentication fails with global proxy enabled - CodexUpgrade cc-switch to v3.16.4+ and enter a working proxy address under 「设置 → 路由 → 全局出站代理」, then retry Codex/ChatGPT authentication.Select Model Menu Still Shows Pre-Switch Model List - CodexClose existing Codex sessions, run `codex app-server daemon restart`, and upgrade CC Switch to v4.0.7 or later for detection and restart prompts.Tray menu does not show current Plan allowance when multiple ChatGPT accounts are linked - CodexUpgrade cc-switch: the tray menu failing to show allowance with multiple linked ChatGPT accounts is a confirmed bug, fixed in commit e53ab21 and released…Error running remote compact task error - CodexUpgrade cc-switch to v3.12.0 or later. It adds forwarding support for the Codex `/responses/compact` route in the local proxy and can resolve 404 or 502 errors.invalid tool_call_id (HTTP 400) — CodexIf tool calls fail, use a native Responses-compatible GPT model; v3.16.5 can be rolled back to v3.16.4.Configuration UI FlickeringUpgrade cc-switch to fix flickering from async form loading; read SQLite DB to fix env vars not updating with providers.auth.json does not change with the channel; editing overwrites the third-party channel key (cc-switch)cc-switch 3.20.0 writes the official key from auth.json back into a Codex relay channel when editing it. Upgrade to the version with the #6534 fix and re-save…The apply_patch tool is unavailable with the Codex proxy onUpgrade to cc-switch v3.16.2 or newer. On a relay, make sure the backend model matches what Codex expects or tool calls fail with routing on.Provider configuration issue, CodexAfter switching providers, restart Codex, then try opening the original session; the configuration writing method changed significantly in v4.0.Wrong model called / stop: cancel task, CodexAfter upgrading to 4.0 or the latest version, try again. Version 4.0 rewrote how Codex configuration is written to prevent the wrong model from being called…Your input exceeds the context window errorThe session context is too long for the model limit. Compact the context by hand or set an auto compact window.ERR_BAD_REQUEST api.anthropic.com ClaudeTurn on "skip the first-install confirmation" in cc-switch and clear the conflicting environment variables to resolve the Claude Code ERR_BAD_REQUEST.Image generation is not enabled for thisManually filter image generation requests or configure a local bridge process to read response headers to work around the 403 block.Minimax configuration overwrittenUpdate cc-switch to the fixed version to resolve the issue where Minimax configuration is incorrectly overwritten during switching.unexpected status 403 ForbiddenDisable 'Keep Official Login' in CC Switch and restart to stop Codex from sending image gen requests causing 403 errors.I'm unable to call create_fileCheck and correctly set the upstream format in SWITCH; an incorrect format configuration made the tools unavailable.Switching Claude relay sites loses previous setting configuration — Claude CodeNo verifiable solution was provided; this request is tracked in main issue #3631.missing field `call_id`Clean up orphaned heartbeat output entries when enabling requests through a proxy; this issue did not confirm that CC Switch has a fix.Added model still shows only Astra: CodexMost related issues remain unresolved. Upgrade to the latest cc-switch and save the model mapping again. If only Astra is shown, provide the version… (Codex)Unable to Change Models After Switching Relay When Quota Runs Out - CodexMost related issues remain unresolved. Retest after upgrading to the latest version. If the problem persists, provide the Codex version, a redacted config.toml…Buttons All Grayed Out and Cannot Send: Codex DesktopNo confirmed fix. Verify the request path and protocol, enable local routing, and test a new session; official quota status may still limit Desktop.Old Provider API Key Overwrites Current Configuration During Proxy RecoveryUpgrade to v4.0.0 or later; the redesigned local routing no longer restores keys from old backups.Form content is overwritten by the last entry when editing, cc-switchThis was a fixed cc-switch frontend issue; the original post did not state the specific version containing the fix.Local proxy port failure: Codex routing errorAfter upgrading to 4.0, sync the configuration and switch to routing mode so CC Switch generates a Codex address using the local listening port.Claude Code verification required when entering ClaudeIn Windows WSL, write the rule that skips Claude Code’s initial confirmation to the corresponding JSON; the maintainer confirmed the issue was fixed.
Anthropic (official)39
Codex auth.json overwritten, KEY mix-up, or official login lostUpgrade cc-switch to the latest version (v3.16.1 fixes the v3.16.0 KEY mix-up regression; v3.20.1/4.0 fix logout overwriting auth.json and losing… (Codex)settings.json config lost after switching providersUse the "Shared Config Snippet" feature to extract and preserve custom fields like hooks/statusLine so they are carried over automatically on switch.Work browser access requires an authenticated accountLog in to the official account in the Codex/ChatGPT app with auth_mode=chatgpt, then use cc-switch routing mode (or keep official login in direct mode) to…Upstream Chat Completions Streaming Response Error, Claude CodeCheck whether the issue is caused by local proxy conversion or the upstream endpoint; adjust the endpoint or disable the local proxy for the verified scenario.401 Invalid Authentication - Claude CodeCheck the API key and authentication variables required by your provider, and confirm Claude Code's actual authentication field contains the correct key.Model list is empty; Codex cannot select the DeepSeek modelUpgrade to CC Switch 4.0 and reapply the DeepSeek preset; model list display in the Codex desktop app also depends on its version support.Claude Code /context Shows 200K Instead of 1MIn Claude Code, run /model claude-opus-4-8[1m], or set ANTHROPIC_DEFAULT_OPUS_MODEL to claude-opus-4-8[1m] in ~/.claude/settings.json.Claude still asks to log in after configuration - Claude CodeAdd hasCompletedOnboarding: true to ~/.claude.json to skip the login prompt.Claude Official check failed: could not extract base_url - CodexCC Switch proxying and load balancing do not support an official Claude subscription — use a third-party API key. Forcing it errors and risks a ban.auth.json not updated after switching channelCodex authentication mechanism refactor causes switching to fail. Check auth.json path and configuration per the official authentication retention guide.The cc-switch Claude config is overwritten - Claude CodeSign out of the official account in cc-switch and clear the login state to restore your developer mode proxy config.The 'glm-5.3' model is not supported - CodexUpgrade to the latest version and disable proxy takeover. In 4.0.5 aggregation mode, start a new chat if old chats still fail randomly.Unknown parameterTurn the cc-switch unified session history switch off and move sessions tagged custom back to openai, fixing the namespace parameter error.Config error: Claude Provider is missing base_url configurationTurn cc-switch local proxy takeover off and run /logout then /login in the terminal, or upgrade past v3.14.0.built-in provider IDs: openai - CodexUpgrade CC Switch to v3.16.4+, delete the old config and save again. Enable unified Codex session history and use openai_base_url.Official account balance information suddenly missing in CodexUpgrade to the latest version, which reworked quota display in 4.0, and check again. If it is still missing, provide the version, login method, plan type…invalid_authentication_error Test Failure HTTP 401An unreplaced {{apiKey}} variable in the usage query script causes 401. Upgrade cc-switch to the latest version; temporarily enter the Key as a literal.HTTP 401 Unauthorized - CC Switch AnthropicUpgrade cc-switch to a fixed build so it sends the x-api-key and anthropic-version headers Anthropic requires.usage null - Claude Code Qwen DashScopeUpgrade to cc-switch v3.16.3+, which fixes null streaming usage on the OpenAI-compatible interface automatically.Chat history under the original codex project disappears after connecting a third party - CodexUpgrade to cc-switch v3.16.4 or newer, turn on "unified Codex session history" in settings, and migrate the official sessions.“导入 0 条, deferred 27–31 个” — Codex Usage Is Not CountedIf it persists after upgrading, wait for the paginated rollout index fix. Disabling auto-sync stops repeated full reads but does not restore statistics.ERR_BAD_REQUEST api.anthropic.com ClaudeTurn on "skip the first-install confirmation" in cc-switch and clear the conflicting environment variables to resolve the Claude Code ERR_BAD_REQUEST.CC switch Does Not Support Multiple Claude Official Logins...CC-Switch cannot switch between multiple official Claude subscriptions. You must log in again; support for this remains under discussion.Unexpected status 503 Service UnavailableWhen using v2ray, configure its routing to bypass LAN and mainland China domains, then access through CC Switch local routing again.auth becomes {}, API Key lost - cc-switch CodexIf ~/.codex/auth.json is missing, saving the active Codex provider clears auth.OPENAI_API_KEY in the DB. Restore the key before saving or avoid saving.The auto mode classifier is temporarily unavailableDisable the Anthropic Base URL proxy in CC Switch and restart Claude Code; one user in the issue reported that updating CLIProxyAPI resolved the problem.Open sessions continue using a third-party ANTHROPIC_BASE_URLSet legacy ANTHROPIC_* keys to empty strings in the official provider configuration's env, then switch back to the official provider.Does Automatic Claude Subscription Balance Refresh Log Proxy IPs?This issue does not say whether automatic refresh records a proxy IP; do not use it to determine which IP the account used.Official account not displaying normally in the codex lower-left corner - CodexThe original post provides no verifiable solution; follow the main issue #3566 referenced by the maintainer.auth.json Is Force-Replaced, Making Codex UnusableUpgrade to a later version that includes the fix. Before upgrading, disable Codex takeover, sign in to official ChatGPT, then re-enable takeover.Usage Limit reached — Codex third-party provider quota exhaustedThis issue report says setting `requires_openai_auth` to `false` for a third-party provider may prevent Codex from checking official account quota and… (API)Configured models not visible - CodexSign in to Codex with an official OpenAI account, retain the official login when switching providers, resave, and restart. Some API Key users report no fix.Only the Last Logged-in Codex Account Is RetainedAuthentication for multiple official accounts in the same ChatGPT Team overwrites each other. No verifiable fix was provided; follow main issue #2245.maximum context length is 1048565 tokens.Reduce PDF or message input. If the DeepSeek-compatible endpoint does not support PDF, stop uploading it, update CC Switch, and retry.Still counted toward GPT official account quota; Codex keeps official loginWith official login retained, official plugins and remote features still run through the official account; the official side counts their usage.Switch back to your official CodeX account and use its quotaAfter adding the official provider, sign out in the Visual Studio Code Codex extension settings and sign in again.Usage query failed cc-switchThird-party APIs require you to adapt the query script yourself; Claude officially does not support usage queries.Usage configuration disappears after a while (cc-switch)Upgrade to v3.16.3; this version retries transient query failures and retains the last successful result.Local proxy port failure: Codex routing errorAfter upgrading to 4.0, sync the configuration and switch to routing mode so CC Switch generates a Codex address using the local listening port.
MiniMax37
truncation_policy.limit hardcoded to 10000 limits CodexIn cc-switch, manually set truncation_policy.limit in cc-switch-model-catalog.json to the model's actual context length as a temporary fix.401 Invalid Authentication - Claude CodeCheck the API key and authentication variables required by your provider, and confirm Claude Code's actual authentication field contains the correct key.Codex conversations with third-party models frequently stop automaticallycc-switch loses function call name/call_id during Responses API conversion. Connect DeepSeek directly to Codex, or upgrade cc-switch and retest.Provider API Key Overwritten: After Codex Restart or SwitchUpgrade to cc-switch v4.0.0 or later to prevent provider keys from overwriting each other and restore the official login state.Check failed Connection failedToggle or switch the system proxy, then fully restart cc-switch. If needed, upgrade macOS or update to v3.16.3+, preferably 4.x.MiniMax usage not showing; test script reports “no result returned” (cc-switch)MiniMax changed its balance endpoint, breaking the old parser. Upgrade to cc-switch v3.16.2 or later; older versions can temporarily use a custom query…auto mode cannot determine the safety: Claude Code errorUpgrade CC Switch first. If the error persists, check thinking parameters, model mapping, stop support, and the response format when stream is omitted.Qwen model errors or truncation - CC SwitchUpgrade to CC Switch v3.16.4 and use the matching new Qwen / OpenCode Go presets.Model list is empty; Codex cannot select the DeepSeek modelUpgrade to CC Switch 4.0 and reapply the DeepSeek preset; model list display in the Codex desktop app also depends on its version support.Add CoderPlan provider preset for Claude CodeFeature request: add a CoderPlan provider preset to cc-switch. A Claude Code preset PR exists, but merge status is unconfirmed.Missing API key.：Claude Desktop OpenCode GoChange the upstream authentication field from ANTHROPIC_AUTH_TOKEN to ANTHROPIC_API_KEY so requests are sent through x-api-key.got 0 from 0 output items: Codex remote compaction failedUpgrade CC Switch to v4.0.0 or later. If you cannot upgrade yet, disable Remote Compaction and use Codex local compaction.DeepSeek usage statistics show the wrong cost / cache hit rateUpgrade to v3.16.3+, fixing input double-counting on the chat→message conversion path, streaming without usage, and wrong model billing attribution.Invalid type for 'messages'Upgrade cc-switch to the latest version for the conversion fix, or downgrade the Codex CLI to 0.80.0 and use wire_api="chat" to work around it.unexpected status 404 Not FoundUpgrade cc-switch to v3.16.4+ and set the upstream format explicitly to Chat Completions on the Codex provider form.Failed to authenticate: Claude Desktop gateway token invalid — ClaudeSign out of the stale official account session in Claude Desktop and restart the app. If the error persists, switch to the default provider and then switch…Only 258k context in codex - CodexOnce cc-switch maps a minimax model onto gpt-5.5, the hard-coded Codex and Claude Code context limits apply and the 1M context is lost.Calling GLM-5 reports error 1311 - cc-switchA provider permissions problem: enable GLM-5 / MiniMax / Kimi access with your provider, or ask them to confirm your subscription entitlements.ending with a leaked `</t` close-tagCaused by the cc-switch streaming converter not deduplicating and not filtering think tags. Update to a version containing the PR #6319 fix.exceeded retry limit, last status: 429 Too Many RequestsAn incorrect GLM preset URL can cause 429: delete the old GLM provider, create the official preset in cc-switch, confirm it uses a Coding Plan link, and check…The user sent an empty message (Claude Code connected to Kimi — repeated empty messages)No official fix: /export the session and continue in a new one; bypass cc-switch so Claude Code connects directly to Kimi to find which layer sends empty…invalid tool_call_id (HTTP 400) — CodexIf tool calls fail, use a native Responses-compatible GPT model; v3.16.5 can be rolled back to v3.16.4.responses_feature_not_supported: web_searchUpgrade cc-switch to v3.16.5, which disables web_search in the config automatically to work around the Xiaomi MiMo gateway limitation.status=429 - Claude DesktopStreaming requests in CC Switch 3.15.0 skip the model rewrite and produce a 429. There is no official fix yet.Cache hit rate only 40% and extremely slow - Claude CodeDisable the dynamic attribution header in `.claude/settings.json`; if that fails, switch to a model with working caching and monitor the main issue.Why is the model the claude command uses in a plain terminal different from the one in cc-switch?Set the MiniMax you want as the primary model and run claude from a cc-switch terminal. Without takeover on, the mapped name is not a substitution.没有这个模型: Codex calls the wrong model after route switchingUpgrade CC Switch to v3.16.1 or later, then verify the Codex route and configuration switching again.Minimax configuration overwrittenUpdate cc-switch to the fixed version to resolve the issue where Minimax configuration is incorrectly overwritten during switching.There's an issue with the selected modelFor NVIDIA NIM / MiniMax, enable the local proxy and rectifier, or export ANTHROPIC_BASE_URL to the local proxy port.unsupported call: - CodexUpgrade cc-switch to v3.16.4+ to fix tool call name/id overwritten by empty deltas during streaming conversion.unknown model 'ark-code-latest' (2013)Change the MiniMax provider's model ID to `MiniMax-M3` or a model MiniMax actually supports, then save. Do not use Volcano Engine Ark's `ark-code-latest`.mcode session usage import: model stays unknown, latency stays 0, and pricing failsNo confirmed user-side fix is available; update the importer to backfill model and latency from telemetry or session data.failed to parse function argumentsNo verified fix is available. Check Codex session logs for truncated arguments JSON, then report it to the upstream provider or switch providers.invalid function arguments json string: CodexUpgrade cc-switch to 3.16.4 or later, then retest with MiniMax-M3.Only under the OpenCode agentUpdate to v3.16.4 or later to use OpenCode Go presets with API Key-only configuration in Claude Code, Claude Desktop, Codex, and OpenCode.Input should be a valid string: Codex image requestUpgrade CC Switch to v3.20.4; it converts image `detail: "original"` to a value accepted upstream.Unknown coding plan provider - Claude CodeUpgrade to v3.20.0 so the Agent Plan `/api/plan` endpoint is recognized correctly.
GitHub Copilot30
Switching providers resets claude settings.json - Claude CodeBefore switching, extract the current provider’s configuration in “General Configuration” and select “Write to General Configuration” for the new provider.EGL_BAD_PARAMETER: CC Switch White Screen on LinuxFor affected Linux AppImage environments, preload the host libwayland-client first. If the white screen remains, try the .deb package.MiniMax usage not showing; test script reports “no result returned” (cc-switch)MiniMax changed its balance endpoint, breaking the old parser. Upgrade to cc-switch v3.16.2 or later; older versions can temporarily use a custom query…auto mode cannot determine the safety: Claude Code errorUpgrade CC Switch first. If the error persists, check thinking parameters, model mapping, stop support, and the response format when stream is omitted."Not logged in · Please run /login" - Claude CodeUpgrade cc-switch to v3.19.2+, or change ANTHROPIC_API_KEY to ANTHROPIC_AUTH_TOKEN in settings.json by hand to skip the approval prompt.API Error: 400 Param Incorrect Claude CodeUpgrade CC Switch to v3.16.2+ to fix 400 errors caused by missing reasoning_content in MiMo thinking mode.missing field `call_id`: Codex automation HTTP 400No verified published fix is available. Normalize orphaned output without `call_id` during proxy forwarding and retain cross-task message content.HTTP 400 / Error 400Upgrade cc-switch first and verify API format and auth fields per provider; if HTTP 400 persists, fix the request using the exact upstream field. Copilot fix…Gtk-CRITICAL: gtk_widget_get_scale_factor failedOn a GTK assertion failure running cc-switch under Linux or WSL2, edit settings.json to turn silent startup off, or clean the leftovers and reinstall.InputValidationErrorWhen long tool arguments are truncated during streaming SSE, temporarily disable Tool Search, maximum-effort thinking, and the 1M declaration, then restart and…Not logged in ? Please run /loginEnable local proxy takeover for Claude Code, point ANTHROPIC_BASE_URL to http://127.0.0.1:15721, and set proxy-managed authentication placeholders.400 bad requestEnable local routing in CC Switch; other users recovered with the native API format, default authentication fields, and a VPN.Not logged in · Please run /login ClaudeEdit `~/.claude.json` to move `PROXY_MANAGED` into the `approved` list, or answer yes at the prompt.API Error: 400 model_not_supportedRestart CC Switch to repair the proxy routing, or confirm whether the model ID requested became invalid after a Copilot policy change.The API returns a 400 error - Claude CodeUpgrade CC Switch to v3.16.1 for the 400 caused by MiMo missing reasoning_content; on Github Copilot, check the API format and proxy config.Main Window Never Appears; CC Switch Fails to Start on WindowsCheck Windows updates and WebView2 arguments first; the issues do not confirm a general fix for all cases.However, this session isn't eligibleUpdate to a version with the #7893 fix; it writes `CLAUDE_CODE_AUTO_MODE_SERVER` to settings.json when switching proxies.Login required: Claude Code login prompt after takeover with Codex OAuthUser-tested: after Claude takeover, keep ANTHROPIC_AUTH_TOKEN (even empty) in ~/.claude/settings.json env before starting Claude Code to avoid login prompts.Format conversion error: No choices in response - CC SwitchClinePass wraps the response in a data envelope, breaking parsing. Upgrade CC Switch to a version with the fix, or use a local proxy to unwrap response.data.Codex still loads skills from ~/.agents/skills after being disabledCheck the skill storage location. If a copy is in ~/.agents/skills, Codex will keep loading it; disable or remove that copy yourself.Copilot subscription model mapping not applied, still calls Claude modelsThis is a known bug in cc-switch v3.14.1. Upgrading to v3.15.0 fixes the issue where Copilot subscription model mapping is not applied.检测到无限空白 bug: Copilot incorrectly drops tool calls containing indented codeA comment reports a fix in commit 33eafbad5198d9fe64ff507ac673430ec334b4aa. Check whether your version includes it; no fixed release was specified.Sessions may stop responding (JSONL stops being written) — Claude CodeUpdate CC Switch to the latest 4.0 release and retry. If it still hangs, temporarily bypass CC Switch and submit complete reproduction details.Top-level model.provider not updated for Hermes DesktopThis report was closed as a duplicate. No verifiable solution was provided; track root-cause issue #5142.After enabling proxy mode, model mappings are not written to Claude Code settings.jsonLocal Routing does not write model mappings to Claude Code settings.json. Configure them in the current CC Switch Provider and switch to enable them.Agent model changes revert after saving and reopening the configurationThe maintainer said this should be fixed in v3.17.0. Try upgrading, saving again, and checking the Agent model. No user confirmed success in the thread.[DISCUSS] Support for pi coding agentThis is a feature request merged into tracking issue #1855; the original post provides no verifiable solution.CC Switch is currently a black box; users cannot customize processing logicThis is a feature request for programmable routing transformation scripts in CC Switch; the original post has no verifiable official solution.Cannot start normally: window unresponsive, system requires force quitUpdate CC Switch to 4.0 or the latest version and retry. If it persists, provide the full version, steps, logs, and screenshots.Warmup request downgraded to modelSet the Copilot optimizer's `warmupDowngrade` to `false` to prevent warmup requests from being rewritten to the unavailable `gpt-5-mini`.
Qwen / 通义千问27
Codex switching DeepSeek model error or only showing customEnable local routing and ensure Codex is logged into the official account to generate models_cache.json, restoring the model list.DeepSeek Claude Code Configuration and Switching IssuesFirst confirm the issue type: new terminals or sessions reload configuration. For abnormal DeepSeek cache hit rates, try v3.16.3; other variants remain…Upstream Chat Completions Streaming Response Error, Claude CodeCheck whether the issue is caused by local proxy conversion or the upstream endpoint; adjust the endpoint or disable the local proxy for the verified scenario.truncation_policy.limit hardcoded to 10000 limits CodexIn cc-switch, manually set truncation_policy.limit in cc-switch-model-catalog.json to the model's actual context length as a temporary fix.api failed deserialize json body intoDepending on the error: roll Claude Code back, upgrade cc-switch, turn the local proxy off, or stop sending unsupported fields.Codex Kimi HTTP 400 moonshot flavored json schemaUpgrade CC Switch to the latest version; the proxy automatically cleans Kimi-incompatible `$ref` sibling keywords and tool schemas missing root-level `type`.EGL_BAD_PARAMETER: CC Switch White Screen on LinuxFor affected Linux AppImage environments, preload the host libwayland-client first. If the white screen remains, try the .deb package.Provider API Key Overwritten: After Codex Restart or SwitchUpgrade to cc-switch v4.0.0 or later to prevent provider keys from overwriting each other and restore the official login state.Qwen model errors or truncation - CC SwitchUpgrade to CC Switch v3.16.4 and use the matching new Qwen / OpenCode Go presets.Error from provider (Console Go): Upstream request failedCheck whether the model mapping was applied. If the target model is unavailable or the request body contains an unsupported tool_choice or image… (cc-switch)kimi k2 reports 404 - CC SwitchA 404 from the connection test does not mean Kimi is unavailable. Set ANTHROPIC_BASE_URL to https://api.kimi.com/coding/ and verify by running Claude Code.Request body format invalid error - Claude CodeUpgrade cc-switch and enable the Rectifier, or add MAX_THINKING_TOKENS=0, for compatibility with third-party API formats.selected model (kimi-k2.5) may not exist errorRemove the trailing /v1 from the API address and use the Anthropic-compatible interface. If an old session contains images, run /compact or switch models.HTTP 400 / Error 400Upgrade cc-switch first and verify API format and auth fields per provider; if HTTP 400 persists, fix the request using the exact upstream field. Copilot fix…Codex image requests return 400 through proxy: base64 inflation or invalid detailIf image history causes context overflow, upgrade to v3.19.0; if detail original is invalid, disable supports_image_detail_original for the affected model.Unable to Display Model Reasoning: Claude Code ProxyUpgrade to cc-switch 4.0.13 and retest. If the issue persists, ignore late reasoning deltas after the response body begins.usage null - Claude Code Qwen DashScopeUpgrade to cc-switch v3.16.3+, which fixes null streaming usage on the OpenAI-compatible interface automatically.There's an issue with the selected modelRemove the trailing /v1 from ANTHROPIC_BASE_URL, launch from the CLI instead of the cc-switch button to use local routing, and check model support.X-OpenAI-Internal-Codex-Responses-LiteComment out the custom provider model_provider setting and return to the built-in openai provider; alternatively, temporarily switch to gpt-5.4.Invalid pages parameter: "" Claude CodeSwitch the API format to anthropic, or upgrade cc-switch to fix the empty pages field from Responses conversion.No endpoints available that support Anthropic - Claude CodeUpgrade Claude Code to v2.1.92+ and update cc-switch to fix OpenRouter parameter errors, or disable the local proxy.The provided messages input is invalidIf a session keeps failing after triggering a built-in tool, start a new session; check whether the third-party backend rejects Anthropic internal content…400 MissingParameter: input.typeNo verified fix is available. The reporter's direct tests returned 200 with "type": "message", but the proxy patch remains a proposal.Bash(...) Waiting... Claude CodeDowngrade Claude Code to v2.1.104 or v2.0.64 (admin required), or switch to a compatible provider such as Alibaba Cloud Bailian.Tool calls always hang: Codex function_call loses call_idA commenter confirmed non-empty id/name checks restore tool calls. The post also suggests bypassing CC Switch Responses translation to connect directly to…Only under the OpenCode agentUpdate to v3.16.4 or later to use OpenCode Go presets with API Key-only configuration in Claude Code, Claude Desktop, Codex, and OpenCode.image_detail is invalid: Codex StepFun Screenshot Request FailureUpgrade CC Switch to v3.20.4; it changes tool screenshot detail: "original" to the StepFun-accepted "auto".
NVIDIA22
Qwen model errors or truncation - CC SwitchUpgrade to CC Switch v3.16.4 and use the matching new Qwen / OpenCode Go presets.DeepSeek usage statistics show the wrong cost / cache hit rateUpgrade to v3.16.3+, fixing input double-counting on the chat→message conversion path, streaming without usage, and wrong model billing attribution.HTTP 400 / Error 400Upgrade cc-switch first and verify API format and auth fields per provider; if HTTP 400 persists, fix the request using the exact upstream field. Copilot fix…Unsupported parameter(s): prompt_cache_keyUpgrade cc-switch to the latest version, fixing prompt_cache_key not being stripped automatically when forwarding to the Nvidia API.Could not create default EGL displayOn Linux, cc-switch forces WEBKIT_DISABLE_DMABUF_RENDERER=1, causing a blank window. Set it to 0 before launch to restore the UI.Request forwarding failed: error sending request for urlCheck your local network connection and proxy configuration so cc-switch can reach the Nvidia API endpoint, resolving the 502 timeout.API Error: 422 format conversion errorThe upstream NVIDIA NIM GLM-5.2 model intermittently returns an empty choices array. Turn on failover in cc-switch or retry the request.Upstream HTTP 404 cc-switchUpgrade cc-switch to remove ?beta=true from NVIDIA requests. Kimi does not support Codex's Responses API; use Claude Code or a local adapter.Invalid pages parameter: "" Claude CodeSwitch the API format to anthropic, or upgrade cc-switch to fix the empty pages field from Responses conversion.KMS: DRM_IOCTL_MODE_CREATE_DUMB failed: Permission deniedResolve cc-switch white screen DRM permission errors on Linux by disabling WebKit hardware acceleration, specifying NVIDIA rendering, or using the CLI version.There's an issue with the selected modelFor NVIDIA NIM / MiniMax, enable the local proxy and rectifier, or export ANTHROPIC_BASE_URL to the local proxy port.Nvidia check failed: Not found (404)Upgrade cc-switch to v3.16.3+; new version uses lightweight HTTP probing, fixing NVIDIA interface false 404 errors.not found(404) adding Claude Code desktop model APIUpgrade to v3.16.3+ to fix connectivity detection false positives, or ignore the 404 and test actual calls directly.CC Switch flashes once when clicking any buttonOn Windows, use WebView2 152.0.4191.62 Fixed Version only for CC Switch to avoid click flashing with WebView2 153.0.4234.32.Unable to Call Tools Normally with Claude CodeModel responses vary on the NVIDIA channel; try deepseek-ai/deepseek-v4-flash, which was tested as stable, or switch providers.Unexpected status 503 Service UnavailableWhen using v2ray, configure its routing to bypass LAN and mainland China domains, then access through CC Switch local routing again.Bash(...) Waiting... Claude CodeDowngrade Claude Code to v2.1.104 or v2.0.64 (admin required), or switch to a compatible provider such as Alibaba Cloud Bailian.Input field cannot gain focus / cannot enter OpenCodeThere is currently no confirmed fix. In Linux Wayland and NVIDIA environments, the app freezes after entering an edit page.max_tokens must be at least 1, got -1712The original post only suggested a lower-bound check for `max_tokens`; no verified fix or user-usable configuration is available.Tasks cannot run continuously: cc-switch 英伟达 providerThe original post considered this a model issue; cc-switch only manages configuration, and no verifiable general solution is currently available.Still unable to use: opencodeThe original post provides no verifiable solution. Follow Issue #884 for the proxy type and no-request problem.The statistics page does not show these Claude requests at allIf the upstream returns no parseable usage, converted Claude openai_chat/openai_responses requests may not appear in statistics; no verified fix is available.
OpenRouter18
Unknown name "prompt_cache_key" - cc-switchUpgrade cc-switch to v3.16.2+ to fix prompt_cache_key injection, or use an OpenRouter endpoint that natively speaks the Claude Code format.Manually edited MCP configuration files are overwritten by cc-switchEdit MCP settings in cc-switch and upgrade to v3.13.0+; use “Common Configuration” to share settings.json plugins/hooks across providers.Codex unavailable: reconnect failed 5 timesAfter switching providers, verify and re-enter the current provider's API Key; API Key overwrite reports continued in v3.16.3.502 Bad GatewayCodex/Claude Desktop reports 502 due to CLI proxy conflicts or app-server bypassing proxy. Remove proxy config or downgrade versions to fix.unexpected status 404 Not FoundUpgrade cc-switch to v3.16.4+ and set the upstream format explicitly to Chat Completions on the Codex provider form.Unsupported content type: redacted_thinkingFor OpenRouter routes, use a proxy to filter redacted_thinking. Codex/GPT routes have no confirmed fix; upgrade to the latest version and retest.401 Unauthorized: Incorrect API key providedCodex via CC Switch to DeepSeek returns 401; manually add http_headers auth headers (Authorization: Bearer or x-api-key) in the config file.The user sent an empty message (Claude Code connected to Kimi — repeated empty messages)No official fix: /export the session and continue in a new one; bypass cc-switch so Claude Code connects directly to Kimi to find which layer sends empty…status=429 - Claude DesktopStreaming requests in CC Switch 3.15.0 skip the model rewrite and produce a 429. There is no official fix yet.Bash failed due to the following issueSet the local GLM provider to OpenAI Chat format and enable routing so CC Switch converts and forwards Claude Code requests.ccstatusline shows the wrong model name - Claude CodeConfigure a Custom Command script in ccstatusline that reads the environment variables from the cc-switch database to show the real model.No endpoints available that support Anthropic - Claude CodeUpgrade Claude Code to v2.1.92+ and update cc-switch to fix OpenRouter parameter errors, or disable the local proxy.Reasoning Level Not AdaptedRe-add the OpenCode Go preset after upgrading to get the correct reasoning_effort config.thinking enabled but reasoning_content missingDisable Thinking mode in Claude Code, or upgrade cc-switch to fix Kimi/Moonshot tool call errors.Incorrect model displayed in Claude CodeConfigure a Custom Command for ccstatusline to read the model name from the current cc-switch provider configuration.Not logged in · Please run /loginWith the Codex provider, do not start Claude Code from the cc-switch built-in terminal. Open a global terminal so local routing handles OAuth.level "max" not supported, Codex DesktopSelect max for gpt-5.6-sol in Codex Desktop; ultra is not a separate reasoning effort.[DISCUSS] Support for pi coding agentThis is a feature request merged into tracking issue #1855; the original post provides no verifiable solution.
Volcengine Ark / Doubao14
Codex Kimi HTTP 400 moonshot flavored json schemaUpgrade CC Switch to the latest version; the proxy automatically cleans Kimi-incompatible `$ref` sibling keywords and tool schemas missing root-level `type`.Auth conflictLog out of the official Claude login, clear conflicting ANTHROPIC_* variables, set hasCompletedOnboarding to true in ~/.claude.json, and restart the terminal.Unkonw coding plan provider - cc-switch VolcengineUpgrade cc-switch to v3.20.0 or newer, then recreate the Volcengine Agent Plan provider or fix its Base URL.Codex image requests return 400 through proxy: base64 inflation or invalid detailIf image history causes context overflow, upgrade to v3.19.0; if detail original is invalid, disable supports_image_detail_original for the affected model.ModelScope reports 401 token invalid - Claude CodeAdd ANTHROPIC_MODEL to the provider environment variables in cc-switch to resolve the 401. It is not synced to settings.json by design.usage null - Claude Code Qwen DashScopeUpgrade to cc-switch v3.16.3+, which fixes null streaming usage on the OpenAI-compatible interface automatically.Gateway rejected model "claude-haiku-4-5"Enable routing based on the API type; for 火山方舟 Responses API, use the full URL and set `isFullUrl=true` to prevent the path from being appended again.config conflicts with official documentation, and automatic model list retrieval fails - CodexUpdate config.toml to match the official specification and set ARK_API_KEY. Codex works again, but automatic model retrieval still needs CC Switch support.400 MissingParameter: input.typeNo verified fix is available. The reporter's direct tests returned 200 with "type": "message", but the proxy patch remains a proposal.unknown model 'ark-code-latest' (2013)Change the MiniMax provider's model ID to `MiniMax-M3` or a model MiniMax actually supports, then save. Do not use Volcano Engine Ark's `ark-code-latest`.image_detail is invalid: Codex StepFun Screenshot Request FailureUpgrade CC Switch to v3.20.4; it changes tool screenshot detail: "original" to the StepFun-accepted "auto".Unknown coding plan provider - Claude CodeUpgrade to v3.20.0 so the Agent Plan `/api/plan` endpoint is recognized correctly.Unkonw coding plan provider Claude CodeUpgrade to v3.20.0+ and recreate the old Agent Plan provider, or change its Base URL to /api/plan.Changed files cannot show diff below. CodexDo not automatically fetch 火山方舟 Coding Plan models. Manually use deepseek-v4-flash, then restart Codex and CC Switch.
SiliconFlow / 硅基流动7
Codex model catalog template `gpt-5.5`Run `codex debug models --bundled > ~/.codex/models_cache.json` to generate the missing model cache, then switch the Codex route again.Auth conflictLog out of the official Claude login, clear conflicting ANTHROPIC_* variables, set hasCompletedOnboarding to true in ~/.claude.json, and restart the terminal.kimi k2 reports 404 - CC SwitchA 404 from the connection test does not mean Kimi is unavailable. Set ANTHROPIC_BASE_URL to https://api.kimi.com/coding/ and verify by running Claude Code.There's an issue with the selected modelRemove the trailing /v1 from ANTHROPIC_BASE_URL, launch from the CLI instead of the cc-switch button to use local routing, and check model support.thinking enabled but reasoning_content missingDisable Thinking mode in Claude Code, or upgrade cc-switch to fix Kimi/Moonshot tool call errors.thinking type should be enabled or disabledThird-party APIs require thinking param as enabled/disabled. Fix via env var CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING=1 or alt+t toggle.Blank or missing model names in /models — OpenCode v2Remove empty name fields from the OpenCode configuration to restore names. Saving again in CC Switch brings the issue back; no official fix is documented.
ModelScope / 魔搭1
Not provider-specific422
cc-switch installation, update, and startup issue collectionThis issue cluster mixes installation, update, startup, and other failures. The available evidence cannot confirm a single fix for the whole page.There's an issue with the selected modelUpgrade CC Switch, restart the target tool, and start a new session so local routing strips the [1m] suffix and rematches the mapping.cc-switch import / cloud sync failures: foreign keys, read-only database, overwritesUpgrade cc-switch and avoid syncing provider_health or a read-only db; after picking a file, click 导入 (Import) rather than 保存 (Save).Opening cc on Windows 11 gives a process but no window - cc-switchOn Windows 11, try Windows 7 compatibility in the exe properties. If that fails, use cc-switch-cli to bypass the Tauri 2 GUI white screen.'claude' is not recognized as an internal or external commandConfigure the full claude.exe path in CC Switch settings, or ensure claude is on the system PATH.Configuration Not Applied or Lost After Switching ProvidersUpgrade cc-switch to v3.16.2+ (v3.18.0 fixes nvm path detection). Put custom settings in “通用配置” to prevent overwrites; Linux save failures require v3.11.1+.cc-switch Local Proxy FAQ: 502 Bad Gateway, Timeouts, Unable to Delete ConfigurationsProxy issues often stem from unsupported official providers, unapplied settings, or unsynchronized UI state;cc-switch Usage Query Not Showing / Query FailedUpgrade to the latest version; enable official account usage manually, check supplier script endpoints and HTTPS. Balance display issues have been fixed since…cc-switch crashes on launch / white screen / black screen / background crashUpgrade to the latest. Windows white screen: install WebView2. Proxy crash: change port, upgrade to v3.9.1+. Old macOS/Linux: use the CLI or distro package.No MCP servers configured. Claude CodeEnable MCP servers one at a time to find the unusable configuration, then disable or remove it and reload Claude Code.cc-switch usage query fails / keeps showing failed / undefinedUpgrade to v3.16.2+. If it still fails, fill in base_url and the access token in NewAPI format, or turn off auto-refresh and the usage query.'claude' is not an internal or external command, nor a runnable programUpgrade to v3.16.2+ to fix WSL/Windows path detection, or manually configure install directory override.cc-switch Skills management problems (duplicates, not detected, deleted repos returning, inconsistent state)Most Skills problems are fixed in later versions: v3.18.0 stops deleted default repos returning; v3.13.0 improves installed-Skill detection and dedup.cc-switch update check fails / update install failsUsually insufficient permissions or GitHub network trouble. Install as administrator or use the portable build, or download manually from GitHub Releases.macOS cc-switch version detection error / freezeModify the fastfetch call in .zshrc to add a [[ -t 1 ]] check for non-interactive output.All app page elements disappear when adding a provider in cc-switchOn macOS 12.2, cc-switch 3.16.x goes blank when adding a provider; upgrade to macOS 12.7.6 or use cc-switch 4.0+.CC Switch Proxy Fails or Requests Are Blocked by the System ProxySet CC Switch's global outbound proxy to your proxy application's address, or bypass local requests in the system proxy, then restart CC Switch.cc-switch Skills Import/Sync Issues (Invalid, Unrecognized, Out of Sync)Most issues stem from Skills directory changes or a DB/filesystem mismatch. Manage Skills through the cc-switch UI, upgrade, and check nested and plugin…cc-switch installation failure: dependency conflict, insufficient permissions, and antivirus false positivesOn older Linux, use a community build for missing dependencies; on Windows, install to AppData if permissions are insufficient;Config not applied: MCP not imported and environment variables overwrittencc-switch overwrites wholesale. Extract hand-edited config into 通用配置 (general config), import MCP by hand, and tick the write-general-config option.cc-switch provider management: cannot delete, or adding failsMostly UI limits or version bugs. The last provider cannot be deleted; for input and delete failures, upgrade or turn the feature off.MCP stdio connection failure: cmd /c wrapper prevents Claude Code from connectingOn Windows, if editing MCP configuration introduces a cmd /c wrapper and causes connection failure, check the generated command and args, then upgrade to a…cc-switch usage statistics anomalies: 0/200K, double billing, pricing that will not deleteUpgrade to v3.16.3 for zero tokens, v3.17.0 for better Codex derived-session dedup, and v3.19.0 for deleted built-in pricing returning.Codex error after enabling local proxyAdd `/v1` to the end of the Codex provider request URL in CC Switch, save, and retry.Codex usage statistics abnormally high or stopped updatingIdentify the problem type and upgrade to the corresponding fix version; fork double counting was fixed in v3.18.0, Windows file growth missed syncing in…cc-switch: two of the five default MCPs fail to addInstall missing `uvx` and `mcp-server-time` dependencies (`pipx install uv`, `pipx install mcp-server-time`), then re-add the MCP.There's an issue with the selected modelCheck the model name and provider config in cc-switch. The model test is unreliable — go by an actual request.Not logged in · Run /login - Claude CodeTurn CC Switch local proxy takeover on, check the API address includes /v1, or launch the terminal from CC Switch so env vars are injected.Codex Session history costs not recalculated, total cost severely understatedUsage costs are computed once at import, causing understated or duplicated costs: upgrade cc-switch to v3.18.0+ (first launch auto-rebuilds Codex history) and…Claude Code Auth conflict errorRemove primaryApiKey from ~/.claude.json or add "auth_mode": "bearer_only". Restore overwritten settings via default or manage shared settings with Common…cc-switch Codex session history missingDisable Codex routing or unified history in cc-switch, make model_provider match old sessions, then restart Codex to restore history.env: node: No such file or directoryWhen CC Switch is launched from the Dock/Finder, the GUI PATH lacks the Homebrew directory; upgrading to v3.18.0+ fixes it.Usage Stats Missing and Disk Usage at 100%Most related issues remain unresolved. Updating to the latest version and disabling the local proxy may restore usage display and reduce disk I/O.Atomic replace failed — request not supported, os error 50 - cc-switchUpgrade cc-switch to v3.20.0 or newer, which fixes the regression where atomic config file replacement failed on WSL paths.It says 1.6 billion used — usage statistics spiking - CodexUpgrade cc-switch to v3.18.0 to fix the sub-agent history replay double-count bug; it rebuilds the polluted Codex usage data automatically.Codex/OpenCode shows not installedAdd the npm prefix or fnm alias directory to the user PATH environment variable, then restart CC Switch.All plugins show disabled: Claude CodeUpgrade to CC Switch v4.0.0 or later; it replaces only routing fields and preserves plugin configuration."There's an issue with the selected model"Upgrade cc-switch to v3.13.0+ for the environment variable sync fix, and use the alias mapping variables to override the default model.either event handler is re-entrantOn Windows, try WebView2 parameters if enterprise security software conflicts; the reports provide no generally verified fix.web_search field deleted: cc-switch rewrites Codex config and manual changes are lostcc-switch rewrites Codex config.toml/auth.json from its database on switch or launch, wiping unmanaged manual fields; upgrade and use "Common Config"… (AGENTS)INVALID_API_KEY after switching providers in CodexUpgrade to v3.20.1 or later; third-party switching no longer writes to `auth.json`, preventing provider key/authentication mismatches.Installation reports missing libwebkit2gtk-4.1.so.0 cc-switchOn Rocky 9, the RPM installation fails because a webkit dependency is missing. Upgrade the system to Rocky 10.1 to resolve it.Troubleshooting cc-switch Upgrade Failures, Freezes, and Display CorruptionUpgrade cc-switch: v3.16.3+ fixes update freezes, v3.20.0+ detects winget-installed Claude Code, and v4.0 refactors upgrades. Upgrade winget installs manually.apikey无权限或失效 - Claude Code Model List RetrievalUse curl to verify `/v1/models` access. Remove spaces or a `Bearer ` prefix from the Key; if the endpoint is unavailable, enter the model name manually.Common Issues After Upgrading cc-switchUpgrade to v4.0 or later; most sync, switching, and display issues from older versions are fixed.WSL path configured but skills are not read - cc-switchWindows cc-switch cannot read or install WSL ~/.agents/skills. Install and run the Linux build of cc-switch inside WSL instead.API Error: 404 - Claude Code path assembly or unregistered routeCheck whether the API address was assembled wrongly, or upgrade cc-switch for the 404 from OpenAI-compatible mode and the missing local route.cc-switch login and auth config cannot be entered or deletedUpgrade to v3.17.0 or newer: the Google OAuth misdetection and the legacy-provider gating that hid the API key field are both fixed.System folder picker opens on every click cc-switchRemove the pickDirectory() call from handleOpenTerminal in src/App.tsx to stop the folder picker from opening whenever you click to open a terminal.cc-switch window display anomalies - UI rendering problemsUpgrade to a version with the UI fixes. Address bar overflow and rendering problems are fixed; the title bar background colour awaits a later fix.Main Window Not Visible; Process Appears Only in Task ManagerUpgrade to 4.0 or the latest version. If no interface appears, report it with the system version, reproduction steps, logs, and screenshots.Claude Code cannot see a newly enabled MCP - cc-switchUpgrade cc-switch to the latest version, or copy the contents of the sibling .claude.json into the one inside CLAUDE_CONFIG_DIR by hand.zsh:1: command not found: claudeUpgrade cc-switch to v3.16.4 or newer, fixing PATH not loading when the terminal launches, or switch to the system default terminal for now.settings.json overwritten by cc-switchWrite non-env custom settings to “General Configuration” and select “Write General Configuration”; field-level merging is not planned.General config ignored - cc-switchUpgrade to cc-switch v3.12.2+, where general config fragments persist across a restart. As a stopgap, turn general config off and write the JSON directly.env: node: No such file or directoryUpgrading cc-switch to v3.18.0 or newer fixes the environment variable resolution failure caused by a stripped PATH in the GUI.Adding a provider fails: crypto.randomUUID is not a functionUpgrade to cc-switch v3.9 or newer, which adds a crypto.randomUUID fallback for older macOS; or upgrade macOS past 12.3 Monterey / 11.6.5 Big Sur.Cannot import providers, cannot manage sessions - cc-switchUpgrade cc-switch for jsonc config support. Under WSL2, clone the database to a local Windows path on a schedule to work around SQLite locking.not installed or not executable - OpenCodeUpgrade CC Switch to v3.11.0+ for the path detection fix, or symlink opencode into ~/.local/bin by hand.Config changes in the GUI have no effect at allOMO v4.19.3 moved config to ~/.omo/omo.jsonc, so cc-switch writes to a dead path. Edit the new config file by hand, or upgrade cc-switch.skills synced to apps I did not select - OpenCodeUpgrade cc-switch to v3.12.3 for the implicit inference bug fix, and turn off reading skills from the .claude directory in OpenCode settings.Skill not loaded correctly - cc-switchcc-switch only recognizes directories containing SKILL.md; manually extract the skills subdirectory from marketplace structures for loading.Skill source directory missing SKILL.md; Claude Code skill installation failsIf installation reports missing SKILL.md, check the repository structure and confirm SKILL.md is in the resolved skill directory.CC Switch does not show the remaining quota (Codex free account, Grok Build usage undercount)Codex free accounts cannot query quota due to official API limits, with no direct fix yet; Grok Build usage undercounting requires a source change to streaming…I clicked both the msi and the exe and nothing happensLaunch the installer as administrator from PowerShell. If it still fails, uninstall the old version first, or use the portable build.assertion failedUpgrade cc-switch to the latest version for the underlying GUI framework crash fix, or try closing proxy software and resetting the network.skills will not fetch / plugins rolled back - cc-switchUpgrade cc-switch to v3.9.0 for the skills fetch fix. If plugin config is overwritten, back enablePlugins up into a general config fragment.exceeds model's maximum context length (Claude Code)Set CLAUDE_CODE_AUTO_COMPACT_WINDOW in the Claude Code provider’s advanced settings in cc-switch to a value below the model’s context limit.Model unavailable in this region (Claude Code)Enable “skip initial installation confirmation/do not re-verify login” in CC Switch and restart the terminal. If it still fails, check for an… (URL)API Error: 400 invalid request: Claude Desktop model route is not configuredUpgrade cc-switch to v3.16.0+ and configure routing rules explicitly for sub-task models such as Haiku and Opus on the provider form.Missing environment variableDelete the env_key entry from config.toml, or set OPENAI_API_KEY in Windows environment variables, and confirm the provider supports the native response format.Conversations gone after switching to another provider in CodexSwitch back to the provider used when the sessions were created to view their history; there is no confirmed fix for syncing sessions across providers.Format conversion error: Failed to parse upstream response - Claude CodeUpgrade cc-switch for the streaming response parse fix, or enable full request path and complete the URL for non-standard APIs such as iFlytek.`GLIBC_2.35' not foundNo confirmed compatible release exists from v3.10.1–v3.16.1. Try v4; if it still fails, report the version and reproduction steps.API Error: 400 invalid_parameter_errorCheck the base url and model config, enable the rectifier, or ignore the false 400 from Stream Check.MCP configured and Claude code selected, but not configured in ClaudeCodeUpgrade to the latest CC Switch 4.0 and sync again. If it still fails, save the MCP configuration once and provide logs.The Add Provider button disappeared in cc-switchMaximize or widen the window. If the button remains obscured in an older version, upgrade to v3.19.2 or v4.0.6.Missing API key · Run /login - Claude CodeClear the conflicting ANTHROPIC_BASE_URL environment variable, or upgrade CC Switch to use its built-in environment variable conflict detection.openclaw.json is not valid JSON - OpenClawAdd double quotes around the models key in ~/.openclaw/openclaw.json by hand to fix the strict JSON parse failure.When the key is `limit`, the input field clears automatically in OpenCodeBecause the cc-switch interface treats `limit` as reserved, edit opencode.json directly to configure model context and output limits.protocol handler not registeredOn Linux, change Exec to cc-switch %U in the CC Switch.desktop file so Chromium browsers recognise the import protocol. For WSL sync, check admin rights.cc-switch on Linux freezes when clicking input fields, UI does not refreshUpgrade to cc-switch v3.13.0 or later. It sets WEBKIT_DISABLE_COMPOSITING_MODE=1 at startup and applies a window focus fix.IO 错误: 当文件已存在时，无法创建该文件 (os error 183)A Junction/hard link configuration directory causes the error. Use a regular directory temporarily and wait for a cc-switch fix.Error 2503: CC Switch MSI Installation Fails on Windows 11Open Command Prompt as an administrator and run msiexec /i with the MSI file path.No skill found in the ZIP file (it must contain a SKILL.md)Upgrade cc-switch to v3.16.2 or newer, and delete the old directory by hand before importing a Skill of the same name.Login required: Claude Code missing ANTHROPIC_AUTH_TOKENUpgrade to v3.16.3; on older versions add an ANTHROPIC_AUTH_TOKEN placeholder to Claude Code's env.API Error: 400 Invalid request ErrorRoll the Claude Code CLI or VSCode plugin back to an older version (2.1.150 or 145) and disable auto-update to resolve the 400.Symlink target not found: .DirIcon broken symlink causes - CC SwitchAppImage install fails due to broken .DirIcon symlink. Fix by converting it to a relative path.ASProxy64.dll Access Violation Causes CrashAdd cc-switch.exe to Astrill’s Help → Hard excluded apps, then sign in to Astrill again.Save keeps failing and asks for the API endpointUpgrade to cc-switch v3.11.0 or later. It natively supports AWS Bedrock and no longer requires an API endpoint or API Key.Agents panel still retains the session entry (status shown as stale) - Claude CodeUpdate to the latest CC Switch, delete the session again, and check. If the ghost entry remains, add the version, reproduction steps, and logs, then reopen…Routing State Switch Failed: Claude Configuration File Does Not ExistIf .claude is missing, run Claude Code once to generate the configuration before enabling routing. Upgrade to v3.16.2 if the message only shows {{detail}}.cc-switch token count is 0An older database missing pricing for a new model makes token statistics read 0. Upgrade to a version containing the fix.The interface is completely blank - cc-switchUpgrade cc-switch to the latest version, which disables DMABUF rendering by default, or use the CLI build to resolve the Tauri 2 white screen.Local proxy toggle UI component is truncated and cannot be fully displayedUpgrade cc-switch to the latest version to fix UI truncation caused by window scaling and tray menu overflow.Automatic upgrade fails; reinstalling still shows old Codex versionNo verified solution yet. #3156 suggests uninstalling and reinstalling; #7080's version-detection issue remains unconfirmed.Cache stuck at 13824: Claude CodeUpdate Claude Code from v2.1.148 to v2.1.215 and retry; user testing confirmed the issue no longer occurred.Upstream HTTP 502 - CC Switch (Claude CodeCheck your local network proxy (Clash/ZeroTier) config and make sure it does not interfere with direct requests.effortLevel: max is not supportedSet the CLAUDE_CODE_EFFORT_LEVEL environment variable to high, or to max on v3.14.1+, avoiding the invalid max value mapping.Prompts /login and cannot use Codex API - Claude CodeDo not encode reasoning level in the model ID. Keep gpt-5.4; CC Switch maps it automatically.Configuration overwritten, Claude Code manual settings lostRead “Why did my plugin configuration disappear?” in the cc-switch README. If using v4.0, upgrade and provide the version and reproduction steps if the issue…Claude Code ignores the model switch in an old windowA Claude Code session already open does not pick up a model switch in cc-switch. Close and reopen the window, or use the new model in a new window.Opening a terminal does not use the complete configurationExtract through the "shared configuration" panel and tick write, so a full overwrite does not lose plugins and other shared config.All installed skills lost - Claude CodeUpgrade to v3.13.0 or later and verify; if skills are still lost after restarting Claude Code, reopen the issue with version and reproduction info.The mcp I configured in .claude.json is reset as soon as the proxy startsCC Switch syncs by full overwrite. Import your MCP servers and manage them all inside CC Switch rather than editing .claude.json.Config file sync delay: CC Switch cloud sync directoryAfter upgrading to 4.0, try built-in WebDAV / S3 cloud sync. If you still use a third-party cloud drive directory, retest and report the drive, version, and…Codex Subscription and API Quotas Used SimultaneouslyUpdate CC Switch and Codex to the latest versions; version 4.0 rewrote Codex provider switching to prevent official login credentials from being sent in…Cannot restore from existing provider auth: Codex auth.json missingUpgrade to v3.13.0 if local auth.json is missing but provider auth exists. Empty auth in sync or SQL exports has no verified fix.auth.json reuses the previous provider’s key, CodexUpgrade to cc-switch v4.0.6 or later. Codex switching logic will prevent the new provider from reusing the previous provider’s key.Automatic approval review failedThe CC Switch local proxy received an upstream 403 while handling the Codex /responses endpoint, causing automatic review to fail; no official fix is available.After setting CODEX_HOME, Codex saves configuration to the wrong locationNo verified universal fix was provided. Set the Codex configuration path manually; direct ${xxx} environment-variable reading is unsupported.Codex Desktop model list is empty and cannot be selectedFollow the official login-state instructions in #4420; a user reported that models became selectable again.Codex history disappears and it will not work - CodexUpgrade to cc-switch v3.16.4 to fix zstd decode failures and CODEX_SQLITE_HOME state database probing that lose or desync sessions.Codex reasoning levels reduced to none and highIn the Codex provider editor, select the supported and default reasoning levels in the model row; do not manually edit catalog files that will be regenerated.Routing and system proxy conflict makes Codex unusableSet IPv4 and IPv6 loopback addresses to DIRECT in proxy rule overrides, then retry CC Switch routing.Codex Current Provider Does Not Exist, Cannot Take Over Live Configuration - CodexAdd a custom provider on the Codex page in CC Switch, then enable Codex local routing.Unified Codex Session History Not Working: Sessions Do Not MergeUpgrade to cc-switch v4.0.0 or later and enable “Unified Codex Session History” so official and third-party routes share a history group.codex(session) usage statistics exceed my actual usage - CodexUpgrade CC Switch to v3.19.2, then run “Rebuild Codex Usage” once on the “Usage Statistics” page.Third-party codex api connected, but the official plan is exhausted—CodexUpdate to CC Switch 4.0 or the latest version. If the balance warning remains, provide reproduction details and logs, then reopen the issue.All tool calls return errors: Codex tool call conversion failedUpdate CC Switch to v3.16.4 or later and retry to fix the reported Codex tool call conversion issue.Usage statistics are massively inflated, CodexUpgrade to cc-switch v3.18.0 or later, then rebuild Codex usage on the usage page to fix double counting.Configuration panel jitter, model list flickering wildly: CC SwitchRetest after upgrading to 4.0; if flickering persists, record the affected area, version, and system details, and attach a screen recording.ENOENT: no such file or directory, scandir Claude CodeIn VSCode Remote SSH, configure Claude Code authentication on the remote Linux server and create missing directories; local settings do not sync."Expecting property name enclosed in doubleUpgrade CC Switch to v3.16.4 or newer to fix the JSON parse error caused by a large request body not being decompressed.Failed to initialize gtk backend cc-switchRunning cc-switch in a headless environment (a server or container) errors for lack of GTK. Use the cc-switch-cli command-line build instead.Cannot read the model list when adding a providerThis is a feature request for automatically fetching the model list; the original posts do not provide a verifiable solution.When current usage is 0 and weekly usage is not 0This is a confirmed GLM usage-display bug. The original post said it was fixed, but no verifiable fixed version was provided.Hermes does not take effect; configuration must be copied manuallyOn Windows, upgrade to v3.16.5 or later. This version makes CC Switch write to the configuration directory actually read by Hermes.does not run on macOS versions other than Monterey: CC Switch installation failureRun `brew update` and retry. If the cask is still locked to Monterey, change the macOS condition to `>= :monterey`.Claude Code shows a white screencc-switch may not start on older systems. On Windows, update to 24H2; on macOS, try the maintainer-recommended CC Switch CLI.UI freezes after clicking any input field on LinuxSet WEBKIT_DISABLE_COMPOSITING_MODE=1 or upgrade to v3.14.1 and enable the native title bar to fix Linux UI freezes and unresponsive buttons.Local Environment Check Fails: Claude Code / Codex / OpenCodeOn Windows, upgrade to CC Switch v3.16.1 or later. The Ubuntu report linked only a fix PR and provided no verification result.error code is 2503: CC Switch installation failureWhen error code 2503 occurs, try the NSIS -setup.exe installer; this issue thread does not confirm that the method works.503, Connection Failure, or Timeout - Codex Local RoutingUpgrade to CC Switch 4.0 or the latest version, then re-enable local routing or “Keep Official Login.” If it persists, submit versions, logs, and screenshots.Windows Explorer opens to the "Documents"On Windows, Open Terminal opens Documents instead of a terminal. A suspected cause and suggested fix exist, but no solution has been verified.Request still uses the old session model, CodexSet all relevant provider configurations in the failover queue to the required model; when continuing an old session, select the target model in the Codex…S3 HEAD bucket failure: 404 Not Found (CC Switch)Upgrade to 4.0 or the latest version and retry; older versions append the bucket again when the endpoint already contains it, causing an incorrect request path.cc-switch routing mode overwrites Claude Code settings.jsonAdd custom env vars to CC-Switch Provider JSON, not directly to Claude Code settings.json.Installation failed: Skill directory does not exist cc-switchUpgrade to cc-switch v3.14.0 or use local ZIP installation to fix URL install failures for repos with SKILL.md in the root.Cannot re-import: Claude Code local skillUpgrade to cc-switch 4.0 and retry. If it still fails, provide the version and error details, then reopen the issue.stream disconnected before completion CodexUpgrade cc-switch to v3.14.0 or later to fix connection drops caused by proxy forwarding hop-by-hop response headers.Claude Code opens a folder selection dialogThis is by design for selecting the Claude Code terminal working directory; select the relevant project folder, then open the CLI.Unsupported parameter: 'max_tokens'The route forwards max_tokens to a model that only accepts max_completion_tokens. Verify parameter support or wait for a cc-switch fix.WebDAV MKCOL failed: 400 Bad RequestNo verified fix is available; NAS suggests using PROPFIND to check whether the directory exists after an MKCOL 400.'/d' is not recognized as an internal or external command: Claude CodeOn Windows, first avoid choosing a Chinese working directory; the issue is not confirmed fixed, so verify the actual directory after startup.x-anthropic-billing-header breaks caching - CC SwitchUpgrade CC Switch to v3.15.0+, which fixes the dynamic billing header breaking prefix caching in openai_responses mode.uv trampoline failed to spawn Python childIf this Hermes detection error persists after upgrading to v3.16.1, provide reproduction details; the issue has no verified fix yet.“Homepage Not Responding” — Codex Fails to Add a Blank Configuration FileUpdate CC Switch to the latest version and retry; reproduction details are insufficient, and no fix has been confirmed.ANTHROPIC_AUTH_TOKEN and ANTHROPIC_API_KEYRun `claude daemon stop --any` to stop lingering Claude background processes, then restart and resume the session.ANTHROPIC_BASE_URL locked to localhostAfter updating cc-switch to v3.12.3, the maintainer said this issue was fixed in v3.9.1.Body discarded: Anthropic protocol conversion Claude CodeUpgrade to v3.20.3 or a v4 version containing the fix; the converter ignores trailing empty reasoning_content.Constant 429 errors: Claude Code AnyRouter errorConfirm the model mapping is blank and set "model": "opus[1m]"; if 429 errors continue, the issue may be with AnyRouter upstream, so try another network.stream disconnected before completionUpdate CC Switch to 4.0 or the latest version and retry; the maintainer recommended this, but the original issue did not confirm a resolution.api.tbox.cn is inaccessible - Claude CodeAfter upgrading to v3.20.4, change an existing BaiLing base URL to api.ant-ling.com, or add it again from the preset.Proxy not working, constant reconnect after entering a question; CodexThe separate proxy feature in cc-switch was removed; set the proxy via $env:http_proxy in PowerShell, then run Codex.Unable to switch between different API: Claude CodeUpgrade to cc-switch 4.0; the maintainer said the new version reworks the interface and configuration writing.Toggling proxy or switching third-party accounts invalidates official Codex sign-inNo verifiable solution was provided; the issue was suggested for further reporting to cc-switch-cli.All models mapped to reasoning model (Thinking) value Claude CodeUpgrade CC Switch to v3.13.0 or later to use the fixed proxy-mode model mapping logic.Error in .codex cc-switch.db causes cc-switch startup failureClose cc-switch, clean records with empty base_url in cc-switch.db, or delete the database file and restart.finish_reason: "stop" and no tool_calls, CodexThe original post provides no verifiable solution; the issue was reported only in the cc-switch and Codex Chat Completions conversion chain.“ANTHROPIC_API_KEY or another auth sourceUpgrade CC Switch from 3.16.4 to the refactored 4.0 or later, then retest.Route classifier requests to another provider in Claude CodeThis is an unresolved feature request; the original post provides no verifiable user-side configuration solution.Model automatically switched to haiku cc-switchBuilt-in Claude Code behavior: haiku reads files. cc-switch only manages environment variables and does not change model selection./branch sessions: parent session hard to find in Session ManagerTry claude --resume or claude -r <sessionId> as the proposed workaround. The original post provides no verifiable Session Manager fix.claude 3.20.4 cache remains 0 - Claude CodeIn Claude Code on Windows, cc-switch 3.20.4 shows cache usage and billing as 0; the issue has no verified fix yet.Tool calls in claude cli stream interrupted after configuring a local proxy Claude CodeNo verified solution is available; one user reported switching to claude-code-router as an alternative.messages[1]: missing field `content` (HTTP 400)Upgrade CC Switch to v3.20.4; the converter skips additional_tools carriers without content, preventing strict upstream services from returning 400.Some historical sessions disappeared in CodexIncrease Claude Code's cleanupPeriodDays and use claude --resume to check whether .jsonl sessions remain on disk.Opus Always Shows opus4.8: Claude Code Local RoutingUpgrade to cc-switch 4.0; the maintainer said this version changed how configuration is written but did not confirm that it fixed the issue.Claude Code no longer hot-reloads settings.json after switching providersRestart Claude Code after switching so it reads settings.json again; clash-for-ai can avoid this limitation.timeout error with the API linkThe maintainer suggested trying cc-switch v4.0; if the issue persists after upgrading, provide the version and reproduction steps.Still shows 1M after switching to a non-1M model, Claude CodeRetest on the latest CC Switch; if it still shows 1M and cannot auto-compact, submit the model configuration and actual model ID.API 400 error when using xiaomimimo token with Claude CodeUpgrade to cc-switch v4.0 and retry. If the error persists, provide the version and reproduction steps when reporting it again.Codex usage anomaly: tokens consumed during Claude conversationsClosing Codex App reportedly restored normal usage statistics as a temporary workaround. The root cause and an official fix remain unconfirmed.Private dir leaf redirectsNo verified fix is available in the original report. The error screen suggests restarting from an earlier message, but no success has been reported.1M Context Configuration Not Taking Effect Claude CodeRevert the Opus route in the cc-switch database from claude-opus-4-8 to claude-opus-4-6, then restart Claude Desktop.Very many 500 requests in Claude CodeThe original post has no verified solution. Restore the previous Claude Desktop configuration and follow the source discussion.Configuration overwrites coworkEgressAllowedHosts, making remote Git unavailableUpgrade to CC Switch v3.16.1 or later; the maintainer said the issue was fixed or implemented.WebFetch fails to access external URL - Claude DesktopAdd `"coworkEgressAllowedHosts": ["*"]` to the Claude Desktop 3P gateway profile generated by CC Switch.Cannot configure additional upstream request headers in Claude DesktopThis is an unimplemented feature request; the original post provides no verified solution.Claude Provider Not Found, Unable to Take Over Live Configuration: Claude CodeUpgrade to v4.0 or later; routing logic has been rewritten. Claude Desktop uses model mapping mode, with no manual changes to settings.json needed.Cannot Find the Authorization Login Entry - Claude DesktopOn first run of CC Switch, migrate the provider from Claude Code, then select the official provider automatically added at the bottom of the list.Claude Desktop model selector cannot display the model listThe Claude Desktop model list is controlled by the client. cc-switch local routing cannot make the developer-mode dropdown display models.Provider add failed: Claude Desktop direct provider missing ANTHROPIC_BASE_URLThe original post only reports this error. A comment suggests proxy environment variables, but this is unverified; continue monitoring the source Issue.Settings reset to CC Switch values in Claude DesktopMost related issues remain unresolved. Upgrade to the current main build containing 81df5a08 or a later release, then retry switching providers.Model switching does not take effect, Claude CodeBack up important information, then click `reset` in CC Switch. The maintainer recommends upgrading and trying again after the 4.0 configuration-writing…Remote Control requires claude.aiUpgrade to cc-switch v4.0 or later; switching to Claude Official clears leftover ANTHROPIC_* credentials from settings.json.1-hour cache write cost undercounted by 37.5%: ClaudeUpdate Claude Code to a version containing this fix so 1-hour cache writes are billed at 1.6 times the 5-minute cache write rate.total_cost_usd = 0: Claude CodeAdd pricing for claude-opus-5-5 to the model_pricing seed data to prevent imported request costs from being recorded as $0.Claude Code Skills not detected: plugins directory structure not supportedAdd scanning for unmanaged Skills installed through Claude Code plugins and parse their manifests; this issue remains open, and the fix is unconfirmed.Not logged in · Please run /loginMove PROXY_MANAGED in ~/.claude.json from the rejected approval list to the approved list, then use routing again.OpenAI Chat Completions model not working, Claude CodeRetry after upgrading to 4.0. If the issue persists, provide the provider API format, local routing status, and relevant logs.settings.json Is Fully RewrittenPut retained Claude Code project settings in `.claude/settings.local.json` to avoid loss when cc-switch overwrites user-level `settings.json`.Claude Code preference changes are not synced back during proxy takeoverUpgrade to v4.0.0; when switching providers, Claude Code preferences remain in ~/.claude/settings.json.Claude Code exits whenever the window size changesUpdating the relevant upstream component restored window resizing; the reporter confirmed the upstream update resolved the issue.Shared MCP not working? CodexUpdate to the latest CC Switch and share Claude MCP again. If Codex still cannot see it, provide the version number and the mcp_servers section from…Switching Does Not Take Effect for Claude Code for VS CodeTry enabling “Apply to Claude plugin,” but users in the issue reported that it still did not work. No verifiable solution is available.Cloud Sync Runs Automatically the First Time the App Opens Each DayUpgrade to 4.0 or the latest version and retest. If it persists, reopen the issue with complete steps, logs, or screenshots.Codex CLI 0.156.0 does not refresh custom provider model listThe issue is caused by the Codex client or provider and cannot be fixed by CC Switch; try upgrading to 4.0 and, if it persists, submit the version and…Your access token could not be refreshedLog out and restart the computer, or end the Codex process and sign in again so Codex uses the latest authorization.Unable to switch Codex App to API access modeUpgrade to CC Switch 4.0 or the latest version, then switch to a third-party provider and sign out of Codex App.No Usage Data After Archiving a Task in CodexUpgrade to 4.0 or later, which imports and counts archived Codex session logs.auth.json contamination in CodexIn Codex 3.16.0, the official login state remains in auth.json. Check provider, endpoint, and token for BYOK in config.toml.Codex config import missing auth_modeUpgrade to cc-switch v4.0 or later; when switching to a third-party provider, the Key is written to config.toml, so there is no need to add auth_mode manually.Codex account still appears logged in after signing outThe cc-switch 4.0 interface and configuration writing behavior have changed. If the issue still recurs after upgrading, provide the version and reproduction…会话已过期: Codex authorization refresh fails in Claude CodeUpgrade cc-switch: a Windows user confirmed a fix in 3.15.0. If it persists in v3.16.2 or later, reopen the original issue with details.Codex session import ignores cache_write_input_tokensUpgrade cc-switch or wait for an official fix: Codex session log import misses cache_write_input_tokens, so cache writes are counted as ordinary input and…Unable to Generate Commit Messages - CodexNo verifiable solution was provided. The report describes an upstream streaming connection interruption while generating a commit message.Context compression keeps failing — CodexUpdate CC Switch from 3.17.0 to the latest version and retry. If it still fails, provide system details, reproduction steps, and logs or screenshots.drops non-provider Codex settings: Codex deep-link import drops fieldsUpdate CC Switch to v4.0.13, then re-import the Codex deep link to preserve safe general settings while link parameters manage the connection and key.Default 1M Context Causes 422, CodexUpgrade to v4.0; context window and other model parameters are filled in automatically when adding or pulling a model.Unable to switch reasoning effort and image input in Codexv3.16.5 resets model catalog capabilities to defaults; editing the catalog is temporary and may be overwritten after switching.Codex desktop app has issues, but works without a routing proxyThe issue lacks verifiable resolution steps. Please provide the full error shown in the screenshot and the routing proxy configuration.Still includes `disable_response_storage = true`No verified fix is available. A PR proposes removing disable_response_storage from generation paths, but its merge or release is unconfirmed.Usage statistics fail to record forked child-session usage - CodexNo confirmed fix exists for 3.20.3; wait for cc-switch to fix usage tracking between forked child sessions and the parent rollout.key '$defs' not found in schemaCodex hits tool schema reference validation failures with Grok. No verifiable fix is available in the original report; follow issue #7241 for updates.Local environment check shows “not installed” for Codex and HermesUpgrade to 4.0. The maintainer said many related issues no longer occur after major changes to configuration switching and the interface.Codex keeps consuming tokens even with no sessions or tasksThe Codex idle-usage thread suggests disabling an unnamed feature shown in a screenshot, but provides no verified solution.CCS cannot detect installed CodexUpgrade to the latest version and check Codex on the “Applications” page. If it is still not detected, provide the versions, which codex, codex --version…No API key configured for provider: customUpgrade Codex CLI and CC Switch to the latest versions; CC Switch supports reading the Codex CLI version from config.toml.Keep official login has no effect; Codex shows API loginUpgrade to cc-switch v4.0.0 or later; this version changes how Codex configuration is written to prevent third-party provider keys from replacing official…codex usage statistics still seem incorrect: local usage far below CPANo verifiable fix is available in the original report. The user reports local Codex usage still far below CPA after updating to v3.18.0.Unable to connect: Codex connection issues on macOSNo verifiable solution is available. Specific errors and reproduction steps are needed; the issue's closed status does not prove it was fixed.Disabled MCP takes effect again in CodexThis issue currently has no confirmed fix; if the problem persists, reopen the issue to continue tracking.config.toml mcp_servers configuration save failure in cc-switchThe original report provides no verified solution. Follow up on or reopen the issue.codex startup_timeout_sec cannot be set manually (Codex)Do not use CC Switch to manage Codex MCP instances that require a custom startup_timeout_sec for now.Unsupported content type: redacted_thinkingWhen Claude Code is routed through Codex/GPT, this error has been marked as a duplicate issue; the issue provides no confirmed local fix steps.Model Does Not Support Image Input and Blocks Image Reading - CodexUpgrade CC-Switch to v3.17.0 and resave the provider configuration. If an alias remains text-only, temporarily edit the catalog and report it.Configured service tier `priority`Temporarily editing the model catalog can restore Priority requests, but regeneration overwrites it; no verified permanent solution is available.The free quota has been exhausted. CodexThe report of Codex still using the old model after switching has no verified fix in the original thread. Follow the source discussion for updates.Allow multiple Codex global promptsThis is an unresolved feature request: concatenate multiple enabled Codex global prompts in a stable order.Missing required parameterUpgrade cc-switch to v3.16.4 or later to fix missing function names in Codex native tool calls when routing is enabled.Disable response strange: Codex configuration failureCheck `~/.codex/config.toml` for deprecated `disable_response_storage`, then share the configuration and startup error with maintainers.OpenAI Official shows blocked and cannot switch to official quota — CodexTurn off Codex proxy mode in the upper-left corner of cc-switch. The reporter confirmed this resolved the official provider's blocked status.orphaned [model_providers.*] tableAfter switching to the official Codex OAuth card, no confirmed fix exists; only clean engine-generated provider tables with no remaining references.OpenAI Official usage displayed incorrectlyOpenAI Official usage is zero while Codex (Session) shows a value. No verified solution is available in the issue yet; follow further discussion.No tool output found for function callWait for a CC Switch fix release. Until then, avoid parallel image and other tool calls in one turn to prevent media messages interrupting tool results.Icons are too small, driving perfectionists crazy. CodexThis issue has no confirmed solution; a comment suggests enlarging the icon locally.Codex database routing metadata lost during backfillUpgrade to cc-switch v4.0.0; backfill was removed, and switching now replaces only key fields.Custom configuration overwritten before extraction: Claude CodeIn 4.0, switching providers after upgrading preserves plugins, hooks, permissions, and other content in settings.json.provider_endpoints.url Still Requests the Old AddressAfter changing the Codex provider’s base_url, set provider_endpoints.url to the same new address to restore the model list.Proxy not taking effect; requests go through native codex requests - CodexThe original post has no verifiable solution yet; the issue remains open. Follow Issue #5941 for updates.Common Config disappears during Codex proxyUpgrade to cc-switch 4.0.0 or later; its redesigned routing mode preserves other config.toml content during takeover.Switching between multiple codex accounts shows incorrect quota - CodexThe maintainer confirmed v3.16.5 fixes quota queries using the previous account after switching providers. Upgrade to that version or later and verify again.Usage query shows two results: CodexAfter upgrading to 4.0 or later, check the Codex subscription quota again; if the issue persists, provide a text description, version number, and screenshot.Your access token could not be refreshedUpgrade CC Switch to v3.17.0 or later to fix official ChatGPT OAuth credential issues caused by the takeover route writing a placeholder.5.6sol 1m context change does not take effect - CodexThere is no maintainer-confirmed cause or fix, so it is unclear whether gpt-5.6-sol supports a 1m context.Cannot directly use /v1/responses - CodexExtract `access_token` and `account_id` from `.codex/auth.json`, add them to the request headers, then call the API.Context Window Information Missing When Enabling Routing in CodexCodex loses context window information after routing is enabled; the issue provides no verified fix.Codex Session Usage Not Applying the Default Cost Multiplierv4.0.0 removed the default cost multiplier; after upgrading, adjust the corresponding model price directly in the usage statistics pricing table.Codex cannot display summarized session titles after proxy model useNo cc-switch changes are needed: Codex does not currently support automatic session-title summaries.Settings Reset to Defaults After Restart, CodexUpgrade to CC Switch v4.0.0 or later; when switching providers or using local routing, it replaces only key fields and preserves other settings and plugins.Account does not exist: old UUID CodexUpgrade to v3.20.4, then reselect and save the currently signed-in ChatGPT account in the Codex official provider card.Codex sub-agent created in copy mode incorrectly counts full usage (Codex)Upgrade to v3.18.0; it fixes double-counted historical usage for copy-mode sub-agents and rebuilds session-sourced usage.Invalid schema for functionThe Codex client reports a tool schema validation error. No verifiable fix is available in the original issue yet; follow it for troubleshooting updates.Codex history list can become completelyKeeping `sqlite_home` and migrating the provider bucket in the active SQLite database can restore visible Codex history.There Should Not Be Such a Huge Number of unknown Model Requests — CodexRetest on CC Switch 4.0.13. If many unknown requests remain, reopen the issue with the version, system, reproduction steps, and logs.Codex usage statistics failureAfter Codex updated to 26.917.51856, GPT-6 Sol usage statistics may remain at 0; the issue provides no confirmed fix.[CODEX-SYNC] deferred 4 CodexUpgrade cc-switch to v3.20.3 to fix Windows session files permanently deferred when mtime does not change.Local Codex is up to date, but CC Switch detects an older versionUpgrade to the latest CC Switch, then check the version, install location, and source shown for Codex on the “Applications” page.API still connects to OpenAI - CodexDelete the existing configuration, then re-add and fully configure the provider in cc-switch to generate a new configuration.unexpected status 502 Bad GatewayThe original post shows CC Switch writes `wire_api = "responses"`; no verifiable fix is available.All queries fail: reproducible Claude Code Coding Plan balance queryUpgrade cc-switch from 3.16.1 to v3.16.2 or later, then query the GLM Coding Plan balance again.cache_read=0/128 - Claude CodeNo verified fix exists for cache-read drops under concurrent requests. Keep concurrency at 1 for now and follow the original issue.The generated catalog falls back to `128000`Explicitly set a context window for GPT-5.6 models to avoid falling back to `128000` when none is configured.Expected an ID that begins with 'msg'Create a new task and migrate the summary, or keep the old task on the Chat Completions conversion path; a permanent fix is still unverified.Codex CLI startup error: Codex cross-platform sync of Windows MCP pathsUpdate CC Switch to the latest version and retest. If the error persists, add logs and resubmit; the original issue did not confirm a fix.Clicking “Add” has no response, CodexEnter a non-empty `OPENAI_API_KEY` in `auth.json (JSON)` before adding, then change it back to `{}` afterward; the maintainer plans to remove this validation.Cannot Open Codex or Claude Code After DEB Install on deepin23No verifiable solution is available in the original post. Follow the main discussion in #3133 for repair progress.Unable to import deep link ccswitchClaude Code on Windows v3.16.1 reported being unable to import this deep link; the original post provides no verifiable solution.ANTHROPIC_CUSTOM_HEADERS not saved - Claude CodeUpdate to a version that includes PR #2929 and re-import; the Deep Link's config.env custom fields are preserved.Custom fields lost after saving: configuration rebuilt to default template CodexNo verifiable solution is available. This duplicate report should be followed through the main issue's fix progress.permissions.deny Overwritten Entirely Claude CodeUpgrade to cc-switch v4.0.0 or later; when switching providers, other settings.json content is preserved as is.Deleted preconfigured repository reappears: cc-switch Skill repositoryComments say the issue was fixed by #5356 and merged into `main`; follow the release status of that fix.Deleted provider remains in tray and switching silently failsNo verified solution is available. Residual tray entries and env in ~/.claude.json can make provider switching appear successful but have no effect.Unable to switch providers through CC Switch in Claude DesktopThe issue appears to be on the Claude Desktop client or provider side; CC Switch cannot fix it at its root.Still an api-injected key, CodexAdd and switch to the official provider in preset providers, then sign out of Codex and sign back in with the official account.Every Click Causes an Error in Claude CodeWhen a duplicate key is reported, the original post provides no verifiable fix; check the duplicate key indicated by the message.Editing a provider does not show model ClaudeThis is a known BUG. The maintainer is working on a fix; no verifiable workaround was provided in the original post.Local environment check displays the jDK version, cc-switchThe original post does not yet provide a verifiable solution; this report was marked as a duplicate. Follow the main issue for fix progress.settings.json env not taking effect - Claude CodeClaude Code reads settings.json only at startup. After switching providers, restart Claude Code or set env vars manually.Environment variable conflict dialog cannot be closed - Claude CodeNo verified solution is documented for this issue; the report concerns cc-switch v3.17.0.cc-switch keeps auto-deleting my environment variables OpenClawWhen an environment variable conflict prompt appears, close the window directly; the maintainer states that closing the window does not delete environment…Detected System Environment Variable Conflict; CC Switch Interface ObscuredUpgrade to 4.0 or later. The environment variable conflict notice becomes a page banner with options to dismiss it and view conflicting variables.Not supported model gpt-5.5: Cline automatic failover failedUpgrade to the latest CC Switch 4.0 release and retry. If it still fails, provide the version, system, reproduction steps, and logs.FastMCP tools not visible in tool list: Claude CodeThe issue was traced to the client or provider; no verified fix steps are available in this issue.“获取模型列表失败” Claude CodeUpdate to 4.0 or the latest version and retry. If it still fails, provide the version, system, full reproduction steps, and logs or screenshots.Why doesn’t the fetched model list appear after fetching succeeds? CodexClick “Add model” to create a row first, then click “Fetch model list” and select a model from that row’s dropdown.`meta.isFullUrl` Cannot Be Saved: OpenClaw / OpenCodeNo confirmed fix exists. Reopen the issue with the current version, environment, and reproduction steps, and wait for full URL mode support.Gemini: 无效令牌 after enabling the local proxyTemporarily disable the cc-switch local proxy. Users confirmed Gemini works with it disabled; no verified fix is documented with the proxy enabled.Billing failure: Gemini model costs recorded as 0 in cc-switchAdd the model name after the slash, such as gemini-3-flash-preview, to the pricing list. The reporter confirmed this restores cost tracking.How to check Coding Plan usage? GLM Coding PlanEnter the URL and Authorization from a browser request in cc-switch usage queries, then parse the quota percentage as shown.GLM 5-hour and 7-day remaining quotas are displayed in reversev3.15.0 reports swapped quota displays; the original post provides no verifiable fix.Unable to configure an API obtained from google ai studioFor Gemini, the thread recommends enabling the ccs proxy and testing in the CLI because model checks were unavailable. Claude Code has no verified fix.Cost always 0: GPT-6 Sol usage statistics in CodexEnable “Model Price Sync” in Settings to synchronize pricing data for gpt-6-sol.How to adjust GPT reasoning effort? Claude CodeThe original post did not provide a verifiable adjustment method. Follow subsequent discussion in this issue.No gpt-5.6 Price in CC Switchgpt-5.6 had not been released at the time, so CC Switch had no price available to enter or display.Vendor name displayed as model name, Grok buildConfigure multiple independent models as shown in the issue example, and set the default model and reasoning level in `[models]`.Switching to Grok Official clears config.tomlUpgrade to 4.0. The maintainer said the new version preserves other settings in config.toml.Deep-link import drops multi-model configuration — Grok BuildTo keep multiple relay models in `/model`, manually add the extra `[model.*]` tables to `~/.grok/config.toml`.Bad request (400) Claude CodeUpgrade CC Switch to v3.16.3; the new health check treats HTTP 400 as provider reachability.Hermes environment variables cannot be deleted - Claude CodeCheck Hermes's config.yaml and remove the relevant configuration; this user-comment suggestion was not verified in the original post.Hermes is not detectedThe original report has no verifiable solution; the maintainer merged it into #4162 for tracking.Hermes Agent cannot discover skills in subdirectoriesThis is an unresolved feature request. The original report suggests recursive scanning, but no verified solution is available.Hermes Provider Not Synced to config.yaml - Claude CodeAfter enabling Hermes, copy its provider configuration from the CC Switch database to config.yaml and add the API key and base_url.Hermes Agent still reads its own skillsThe original post provides no verifiable solution. Upgrade and re-project; if it persists, report the actual directories, system, and Hermes version.Conversation terminates without error: Codex generates images with route proxy enabledUpgrade cc-switch to v3.16.3 or the latest version to fix interrupted Codex image requests under local route mapping.ContextWindowExceededError: Codex image uploads exceed the context windowThis issue is unresolved and has no verified fix; it only confirms that Codex sends uploaded images as embedded base64.Local routing failure CodexUpgrade cc-switch from 3.16.1 to 3.16.4 and observe whether the issue recurs; this version includes hardening for Codex routing takeover.JieKou AI users currently need to configureThis is a feature request to add JieKou AI provider presets; the original issue does not yet provide a verifiable user-side solution.Window icon displays the generic “wayland” placeholder icon (orange W)On KDE Wayland, name the installed desktop file cc-switch.desktop so it matches the window app id.Origin Returns 502: Claude Code Large-Context Routing ErrorNo fix is confirmed; provide reproducible conditions for the large context to locate the routing error.Hang When Switching Lightweight Mode, CC SwitchUpgrade to 4.0 or a newer version and check whether the hang still occurs. If it does, provide the triggering action, version, and relevant logs.Unable to Establish Connection (DNS / Connection / TLS / Timeout) Codex LinkfoxRetry configuring Linkfox API after upgrading to cc-switch v4.0. If it still fails, provide the version and reproduction steps when reporting it again.“About” page hangs: Linux cc-switch tool detectionUpgrade to cc-switch 4.0; the maintainer says it substantially changed configuration switching and the interface, and many related problems no longer occur.Clicking “Edit” freezes Claude CodeOn Linux, switch to the deb package, or fix the GTK input-method cache in AppImage to prevent fallback to XIM.Claude routing does not work on Ubuntu — Claude CodeUpgrade to CC Switch 4.0.13 and retry. If it persists, reopen the issue with the version, system, reproduction steps, and relevant logs.Cannot set terminal process group: ioctl operationThe original post has not provided a verified fix. Follow Issue #5635 for updates.Top-right close, maximize, and minimize buttons do not work: cc-switch LinuxOn Wayland, open “Settings → General → Enable app-level window buttons” to use window buttons drawn by the app.Local qwen3-coder model connection fails with 502 errorClaude Code requires Anthropic Messages; use a format-conversion proxy such as LiteLLM between cc-switch and the local model.Local routing uses a hardcoded default model (Claude Code)Upgrade to cc-switch 4.0; the maintainer said many related issues may no longer exist after major configuration changes.How does the routing (proxy) feature work? Claude CodeThe post only states that routing is forwarded through local 127.0.0.1:15721/v1; no verifiable complete configuration solution is available.WebView2 GPU process GPU usage rises abnormally after enabling local routing CodexThis issue was fixed in #5767. Wait for a later release; if it still occurs, reopen the issue and report it.Blank screen when adding a providerUpgrade to CC Switch 4.0.13. If it persists, report it again with the version, system environment, reproduction steps, and logs.Unable to fetch modelsAfter upgrading to v3.16.4, fetch models again. If it still fails, add the provider, error text, or logs to the original issue.Closing to tray saves main window state as 1×1, then reopens tinyDelete or rewrite the abnormal window state file, then restart; PR #7604 fixes hidden windows being saved as 1×1.MacOS Memory Leak: cc-switchRestarting cc-switch after prolonged use can temporarily restore WebKit.WebContent memory usage; no complete fix has been merged.Local custom skills do not update Claude Code after updatesThe original post reports that cc-switch-managed local custom skill updates do not sync to WebDAV; no verifiable solution is provided.MCP Configuration Differences Across CLI Tools, CodexThe same MCP service uses different configuration formats and parameters in Codex and Claude; no verified solution has been provided.Automatic approval review failed: Codex automatic approval with MiMo failedUse a model that supports json_schema for Codex automatic approval, or wait for CC Switch to provide a compatible fallback for MiMo.unexpected end of dataNo verified fix is provided in the original post. Codex receives HTTP 400 and the conversation is interrupted after a MiMo tool call.No auto-detect model option in latest Ubuntu version? cc-switchUpdate cc-switch to v3.12.0 or later; Stream Check is restored and model auto-detection should work on Ubuntu.Mixed LF/CRLF SSE delimiters prevent ClaudeFix SSE delimiter parsing to recognize mixed LF/CRLF blank-line delimiters and preserve Claude Code tool-call events.Model list error: Codex model list does not match the providerMost related issues remain unresolved. Retest after upgrading to the latest Codex configuration-writing logic. If it persists, provide… (/.codex/config.toml)haiku and sonnet model IDs not synced in Claude CodeUpgrade CC Switch to `3.14.1`; it fixes one-click setup not syncing haiku and sonnet model IDs.Model test always returns 404: CodexUpdate to the latest version and test again. If it still returns 404, record the version, provider, API address, and request path from the logs for…Multi-model switching for provider code plans in Claude Code and CodexAfter starting cc, use `/model model-name` to select the model.Effort is silently dropped for all muse-spark-* models in Claude CodeThe issue proposes adding muse-spark to an allowlist, but no verifiable fix is available yet.Direct models must use Sonnet / Opus recognized by Claude DesktopThis is an unresolved feature request: direct-connection validation does not allow the `mythos-` prefix, with no verified solution provided.Failed to fetch models after changing key; connectivity speed test failed - CodexAfter switching internal networks, restart cc-switch to restore the model list and connectivity test; also update to the latest version and retry.Codex model not displayed, CodexThis is a personal reflection on living with BUG; the original post provides no verifiable solution.auth.json Not Taking Effect, Codex Official Channel Cannot AuthenticateUpgrade to 4.0 or later, then retest authentication by switching Codex to the official channel.Codex OpenAI Official cannot participate in failoverOfficial providers are intentionally excluded from failover to avoid account-ban risks because a local proxy cannot fully simulate official requests.Failed to send request: Keep Official Features with a Third-Party Codex ModelNo proxy setup was verified. The maintainer recommends upgrading to cc-switch 4.0 and providing the version and reproduction steps if the problem persists.Multiple cc-switch.db Files When Syncing OneDrive Across DevicesSimultaneous database writes from multiple devices during OneDrive sync can create conflict copies; no verified fix was provided.System message must be at the beginning. in Claude CodeChange Claude Code's "model" from "opus" to "haiku" or another model family to avoid an OpenAI-compatible API 400 response.OpenAI Responses streaming text is duplicated in terminal events Claude CodeTemporarily avoid duplicate output by changing the provider API Format to OpenAI Chat Completions (`openai_chat`).toggling Send User-Agent silently removes every unrelated custom headerDo not toggle Send User-Agent in the OpenClaw provider editor; upgrade and wait for a fix, as the switch replaces or deletes the entire headers object.OpenCode extra-option edits change untouched string values into JSON valuesIn cc-switch 3.20.4, saving an OpenCode provider converts unmodified strings in Extra SDK Options to JSON values. Upgrade to a version with the fix.OMO Configuration Cannot Add Custom subagents - OpenCodeAs a temporary workaround, write all configuration in “Other fields”, but this overwrites the UI configuration above.invalid config structure - OpenCodeFix the update order: save the database row only after the OpenCode live configuration is written successfully; keep the original record if writing fails.OpenCode usage rows never pruned: OpenCodeFix sync logic so the OpenCode session cursor advances despite incomplete messages, preventing historical records from being reinserted after pruning.OpenCode V2 Usage Collection Stops UpdatingUpgrade CC Switch to a version containing the PR #7541 fix so the collector reads OpenCode V2 tables.New-format conversations are invisible in OpenCodeUpgrade CC Switch to 4.0 to read OpenCode 2.x new-format sessions and V1/V2 sessions in the same database.Subagent call error: Claude Code PackyAPIRetest after upgrading to 4.0.1 or later; this version fixed a parameter issue when Claude Code calls subagents through Responses conversion.Duplicate usage records double consumption - PiWhen Pi uses the cc-switch local proxy, the same request may be charged twice; a verified deduplication fix is still pending.Previously selected item remains selected after switching - cc-switch PiAfter switching cc-switch Pi configurations, the previous selection remains retained; the original post provides no verifiable solution.Cannot open the app after Loading, cc-switch PortableThe original post does not provide a verifiable solution; see the issue's follow-up discussion.Model does not respond, Claude CodeDelete all providers, then set the same model for the same provider to restore Claude Code conversations.Fill in the API Key for an unofficial provider - Claude CodeUpgrade to cc-switch 4.0 or later. If the API Key field is still missing, provide the version, provider type, and a screenshot.Model api configuration mix-up CodexThe original post provides no verifiable solution; follow #4108 for fix progress.Codex delete button disabled and greyThis issue has no confirmed fix; in cc-switch 3.20.4, a disabled delete button can keep an incorrect provider configuration active.Search clears, preset list reopens when adding a provider in cc-switchUpgrade to cc-switch 4.0; search for and select a preset before opening the form.Claude Code context full before 256kNo verifiable solution is available. The report shows an error at about 117.9k context and inability to /compact.Proxy Mode Model Mapping Has No Effect - Claude CodeRetry after upgrading to cc-switch 4.0. The maintainer says the refactor fixed the old model-mapping issue, but this thread has no user verification.Usage not tracked Claude CodeAfter enabling the proxy in cc-switch, Claude Code usage is tracked.getTimezoneOffset() returns -8 instead of -480Avoid local-time API on Windows; use timestamp differences or toISOString()/getUTC*, and wait for cc-switch to fix it.Unable to pull complete skill - Claude CodeThe maintainer said the relevant capability had been added, but no version or verifiable steps were provided.Cannot remove all providers — CodexDisable Codex in cc-switch settings. To keep using it, enable OpenAI Official and sign in to Codex again.Cannot run cc-switch after renaming the user folderThe maintainer attributed this issue to the client or provider; no verifiable fix steps were provided.Using a Codex proxy causes multiple Codex requests for one questionUpgrade cc-switch to v3.16.2 or later; the maintainer said it addresses repeated parsing failures and retries.unsupported call: — Codex tool call failedNo confirmed fix yet; CC Switch must preserve function_call name and call_id in streaming events.422 format conversion error: Responses upstreamUpgrade CC Switch to a version containing the #7787 fix: upstream unavailability will be reported as 502 (UpstreamError), rather than wrapped as 422.The editor does not clearly distinguish the “local routing address” from the “third-party provider upstream address”.Configure Claude Code's local routing address separately from the upstream OpenAI-compatible provider address to avoid saving the local address as upstream.Claude Code missing from installed plugin configurationThis report has no verifiable solution. Follow the fix progress in primary issue #3631.model xxx not found: Codex API Key overwrittenUpgrade to a later version containing the #6534 fix to prevent other providers' keys from overwriting the current provider during editing.Session Management Page Narrows When There Are No Records, CC SwitchUpgrade to 4.0; the maintainer says the new session management interface no longer narrows when there are no records.Session expired: still appears after logging in again — CodexThe original post has no verified solution yet; follow the issue discussion if the session still appears expired after logging in again.UI freeze in Sessions ManagementUpdate CC Switch to v3.14.0, which uses session-list virtualization and collapses long messages by default.Save Button Disappears and Directory Is Saved UnexpectedlyMake autosave submit only persisted directory values and preserve unsaved directory drafts during settings refreshes and concurrent saves.Skills cannot be foundThe original post does not yet provide a verified solution; follow the later discussion in Issue #4221.cc-switch: Skill installation says the skill directory does not existWhen installing a Skill on Windows reports that the directory does not exist, no verified fix is available; v3.16.2 did not resolve later reports.Skill description still shows old content in Claude CodeAfter manually editing SKILL.md, Claude Code reads the new content, but cc-switch does not refresh the installation-time metadata snapshot in its skill list.Skill does not exist in SSOT: autoplan - Claude CodeRetest after upgrading to 4.0. If the message still appears, provide autoplan's source, version, and relevant logs.Skill updates silently overwrite local changesBack up local changes before updating; the proposed fix will exclude locally modified Skills from “Update All.”SkillClaw as a shared skill lifecycle layerNot a fault; this integration proposal was closed after prolonged inactivity. If still relevant, add current details and reopen the issue.Skills remain as copies after automatic import, CC SwitchAutomatic import copies Skills to `~/.cc-switch/skills/` and does not delete the source directory in `~/.agents/skills/`.Installed Count Mismatch, Missing Skill Directories, Claude Code SkillsOn the Skills page, check whether the directories for installed entries still exist; remove stale records or rescan the directories before syncing.skills.zip has too many entriesUpgrade the device responsible for downloading or restoring to v3.20.4 and try again; the skills.zip entry limit is now 30,000.skill.sh gstack installation fails - Claude CodeReported on Windows with cc-switch 3.13.0; the maintainer said it was fixed, but gave no verifiable steps or fixed version.Import Stuck for About 30 MinutesUpgrade to CC Switch 4.0.13 and retry the same SQL. If it still hangs, record the file size, wait time, and logs, then report back.Invalid parameterUpgrade to CC Switch 4.0 or the latest version and retry. If it still fails, submit complete reproduction details and logs.baseurl error or 401: Codex's base_url is overridden by the routing addressThe configuration write behavior changed in CC Switch 4.0. If the issue persists after upgrading to 4.0, provide the version and reproduction steps.Codex key abuse after opening CodexThe original report only describes suspected abnormal calls; no verifiable fix or virus file location was provided.Codex history becomes blank after switching configurationRetry after upgrading to CC Switch 4.0 or later; if the issue persists, provide the version and before-and-after model_provider values.429too many requests, CodexA 429 occurs when using a third-party Responses API, but the original post does not provide a verifiable solution.SSH disconnects after network manager removal during installation; Ubuntu fails to bootNo verifiable fix is available. The maintainer said there was insufficient information to reproduce the problem and suggested upgrading to 4.0 and adding the…Configured environment variables seem ineffective, Claude CodeAfter removing cc-switch, manually clean up local configuration as needed; session records in `~/.codex` are unrelated to cc-switch.environment variable cleanup fails - CodexThe original post recommends parsing the line number from the final colon to preserve complete shell configuration file paths containing colons.Self-update check fails and latest model versions show as unknown - Claude CodeUpdate CC Switch to the latest version and retry; related functionality was refactored in 4.0, but the fix was not confirmed.Model does not support text/event-stream, CodexUpgrade to cc-switch v3.16.4 and override the Accept request header in Codex advanced settings with a value accepted by the upstream.Normal requests and Sessions are all counted as CodexUpgrade to v3.15.0 for the confirmed duplicate-counting fix; occasional Codex(Session) entries may still occur.[USG-001] 记录使用量失败: {e}：cc-switchMove the full usage calculation and SQLite write to spawn_blocking, and use a bounded retry queue for failed writes.Claude Code crashes on launchThe original post has not provided a verifiable fix. Follow the subsequent handling in Issue #2923.Environment variables not written: VS Code Claude pluginCheck VS Code `settings.json` for Claude environment variables; if present, comment them out and reopen Claude.vscode Claude Code plugin cannot switch configuration immediately - Claude CodeThe VS Code Claude Code plugin cannot use a proxy or hot-switch; start cc from the terminal CLI for changes to take effect immediately.400 model provider error Claude Code pluginAdd the environment variables required by the current provider to VS Code’s claudeCode.environmentVariables setting so the plugin can read them.dl1-v6.aliyundrive.cloud Claude Code WebDAVNo verifiable fix was provided. The maintainer is tracking this issue under issue #1856.Please check WebDAV username/passwordcc-switch masks and clears WebDAV password on write-back. Upgrade to fix; temporarily re-enter password.Error 2147942402 (0x80070002) — Claude CodeRecheck after upgrading to CC Switch v3.16.1 or later on Windows; temporarily set PowerShell as the preferred terminal.cc-switch keeps opening detection windows when checking local codexWindows CLI detection causes windows and processes to pile up. The issue provides no verifiable fix; a timeout is only a user request.Window Size Overflows to 65535×53251, CC Switch Main Window InvisibleFully exit CC Switch, delete the window state file, and restart to clear the persisted invalid window size.baseUrl: Required Claude Desktop configuration not appliedWindows MSIX Claude Desktop must read 3P configuration from its Packages virtualization directory; the compatibility fix is currently only an submitted PR.Skill does not appear: Claude CodeUpgrade to cc-switch 4.0 and reinstall the skill. If it still fails, add system details and reproduction steps, then reopen the issue.CC Switch Windows opens to a black screen and crashes after a few secondsWait for a later version that disables Overlay on Windows; the original issue provides no verifiable current fix.App Won't Open: cc-switch for WindowsTry installing WebView2 Runtime; an issue comment suggested it, but the original reporter did not confirm whether it worked.Configuration Not Applied: Claude Code WSL Routing Config Not UpdatedSet the configuration directory to the file path used inside WSL in CC Switch Settings > Advanced. If the issue persists after upgrading to 4.0, provide…No models displayed after opening pi - piPoint CC Switch’s Pi configuration directory to the WSL `~/.pi/agent` path. The reporter confirmed that models appeared again.Cannot start cc-switch; cc-switch --version produces no outputThe maintainer recommends reinstalling first; if it still cannot start, try the CLI version. Recovery was not confirmed in the original post.Cannot detect conversation history and usage statistics? Claude CodePoint the Claude configuration directory to the WSL path; route Claude Code in WSL through the cc-switch gateway for usage statistics.3.15.0 hangs on startup under WSLg: CC SwitchRolling back to 3.14.1 restores normal startup; alternatively, reproduce on the latest version and submit WSL, graphics environment, and startup output details.Upstream 404: Claude Code-hosted xAI OAuth Grok billing queryUpgrade to a version containing commit `15d5dbe` to fix the Grok billing query for the hosted xAI OAuth Provider.
This is an unofficial community wiki with no affiliation to the cc-switch authors or the project itself. Its content is compiled from the project's public GitHub issues. This site distributes no software.