A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io
累计 0 次下载请求 · 近 30 天 0 次
查看开发者<div align="center"> # CC Switch ### The All-in-One Manager for Claude Code, Claude Desktop, Codex, Gemini CLI, Grok Build, OpenCode, OpenClaw, Hermes Agent, Pi & MiniMax Code **Switch API providers in one click and manage MCP, Skills, and Prompts in one place — no more hand-editing JSON / TOML / YAML config files.** [](https://github.com/farion1231/cc-switch/releases) [](https://github.com/farion1231/cc-switch/releases) [](https://tauri.app/) [](https://github.com/farion1231/cc-switch/releases/latest) <a href="https://trendshift.io/repositories/15372" target="_blank"><img src="https://trendshift.io/api/badge/repositories/15372" alt="farion1231%2Fcc-switch | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a> <a href="https://www.star-history.com/#farion1231/cc-switch&Date"><picture><source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/badge?repo=farion1231/cc-switch&theme=dark" /><img alt="Star History Rank" src="https://api.star-history.com/badge?repo=farion1231/cc-switch" width="196" height="55" /></picture></a> ### 🌐 The Only Official Website: **[ccswitch.io](https://ccswitch.io)** English | [中文](README_ZH.md) | [日本語](README_JA.md) | [Deutsch](README_DE.md) | [Changelog](CHANGELOG.md) **[Download](#download--installation) · [Quick Start](#quick-start) · [Features](#features) · [FAQ](#faq) · [User Manual](docs/user-manual/en/README.md)** </div> ## ❤️Sponsor > [Want to appear here?](mailto:support@ccswitch.io) <details open> <summary>Click to collapse</summary> [](https://platform.kimi.ai?track_id=track-674ed6e2af924a5682a87421f7cf753a&aff=cc-switch) Kimi K3 is Moonshot AI's most capable model and the world's first open 3T-class model. With 2.8 trillion parameters, native vision, and a 1-million-token context window, K3 delivers frontier performance across long-horizon coding, knowledge work, and reasoning. CC Switch makes it easy to configure and switch to Kimi across agentic tools. Try a **Kimi Code plan** ([中文站](https://www.kimi.com/code?aff=cc-switch) | [Global](https://www.kimi.ai/code?aff=cc-switch)), or use the **API** through the Kimi Open Platform ([中文站](https://platform.kimi.com?track_id=track-7cf2b91dcde043eda6ef9a95951a042c&aff=cc-switch) | [Global](https://platform.kimi.ai?track_id=track-674ed6e2af924a5682a87421f7cf753a&aff=cc-switch)). **New user top-up bonus**: register and complete your first top-up via the API links above to receive 10% of the amount as bonus API credit, up to CNY ¥1,000. --- <table> <tr> <td width="180"><a href="https://www.packyapi.ai/register?aff=cc-switch"><img src="assets/partners/logos/packycode.png" alt="PackyCode" width="150"></a></td> <td>Thanks to PackyCode for sponsoring this project! PackyCode is a reliable and efficient API relay service provider, offering relay services for Claude Code, Codex, Gemini, and more. PackyCode provides special discounts for our software users: register using <a href="https://www.packyapi.ai/register?aff=cc-switch">this link</a> and enter the "cc-switch" promo code during first recharge to get 10% off.</td> </tr> <tr> <td width="180"><a href="https://zetaapi.ai/go/u117"><img src="assets/partners/logos/zetaapi-banner.png" alt="ZetaAPI" width="150"></a></td> <td>Thanks to ZetaAPI for sponsoring this project! ZetaAPI focuses on real model fidelity, no watered-down responses, no quality degradation, and pricing as low as 35% of official rates. The platform does not mix traffic, secretly replace models with lower-quality alternatives, or use fake model routing. It supports Claude Code, Codex, Gemini, ChatGPT, and other mainstream AI models, helping users significantly reduce API costs while maintaining reliable model quality. At the same time, ZetaAPI provides enterprise-grade SLA-backed stability, standard API compatibility, one API key for multiple models, fast integration, and pay-as-you-go billing, making it suitable for AI products, coding agents, internal business tools, customer service systems, content generation, and automation workflows. If any model is verified to be inconsistent with its stated quality, ZetaAPI backs it with a 10x compensation guarantee, giving users a more stable, transparent, and trustworthy experience. Register via <a href="https://zetaapi.ai/go/u117">this link</a> and use the promo code CC-SWITCH during your first recharge to enjoy an exclusive 10% discount on your first top-up, just for CC Switch users!</td> </tr> <tr> <td width="180"><a href="https://apinebula.ai/VjM74M"><img src="assets/partners/logos/apinebula_banner.png" alt="APINebula" width="150"></a></td> <td>Thanks to APINEBULA for sponsoring this project! APINEBULA, an enterprise-grade AI aggregation platform under Galaxy Video Bureau, leverages extensive platform resources to provide developers, teams, and enterprises with stable, cost-effective access to large language model APIs. The platform integrates leading, full-powered models like Claude, GPT, and Gemini, allowing you to connect to the world's top AI models through a single API, with prices starting as low as 10% of the original cost. Designed for AI programming, Agent development, and business system integration, APINEBULA supports enterprise-grade high concurrency, formal contracts, corporate bank transfers, and invoicing services. APINEBULA provides special discounts for our software users: register using <a href="https://apinebula.ai/VjM74M">this link</a> and enter the <strong>"ccswitch"</strong> promo code during your first recharge to get <strong>10% off</strong>.</td> </tr> <tr> <td width="180"><a href="https://www.aicodemirror.ai/register?invitecode=9915W3"><img src="assets/partners/logos/aicodemirror.jpg" alt="AICodeMirror" width="150"></a></td> <td>Thanks to AICodeMirror for sponsoring this project! AICodeMirror provides official high-stability relay services for Claude Code / Codex / Gemini CLI, with enterprise-grade concurrency, fast invoicing, and 24/7 dedicated technical support. Claude Code / Codex / Gemini official channels at 38% / 2% / 9% of original price, with extra discounts on top-ups! AICodeMirror offers special benefits for CC Switch users: register via <a href="https://www.aicodemirror.ai/register?invitecode=9915W3">this link</a> to enjoy 20% off your first top-up, and enterprise customers can get up to 25% off!</td> </tr> <tr> <td width="180"><a href="https://pateway.ai/?ch=etzpm8&aff=WB6M6F67#/"><img src="assets/partners/logos/pateway.png" alt="PatewayAI" width="150"></a></td> <td>Thanks to PatewayAI for sponsoring this project! PatewayAI is an API relay service provider built for heavy AI developers, focused on directly relaying official high-quality model APIs. It offers the full Claude lineup and the Codex series, 100% sourced from official channels — no dilution, no fakes, verification welcome. Billing is transparent and every token-level invoice can be audited line by line. It also supports enterprise-grade concurrency and provides a dedicated management platform for enterprise customers — formal contracts and invoicing are available; visit the official website for contact details. Register now via <a href="https://pateway.ai/?ch=etzpm8&aff=WB6M6F67#/">this link</a> to receive $3 in trial credit. Top-ups go as low as 60% of the original price, with a two-way referral bonus of up to $150!</td> </tr> <tr> <td width="180"><a href="https://api.fenno.ai/register?redirect=/purchase?tab=subscription%26group=16&aff=P9MR3D3PLCNL"><img src="assets/partners/logos/fenno-banner.png" alt="Fenno.ai" width="150"></a></td> <td>Thanks to Fenno.ai for sponsoring this project! Fenno.ai is a stable and efficient API relay service provider, currently focused on Codex relay. It is compatible with the OpenAI and Anthropic protocols and can be flexibly used from Codex, Claude Code, OpenCode, and other mainstream coding tools. It reliably supports enterprise-grade workloads of hundreds of billions of tokens per day, with corporate (B2B) settlement and invoicing for both domestic and overseas entities. Fenno.ai offers an exclusive benefit for CC Switch users: subscribe via <a href="https://api.fenno.ai/register?redirect=/purchase?tab=subscription%26group=16&aff=P9MR3D3PLCNL">this link</a> to the $1.99 Trial Plan worth $50 in credits (valid for 7 days), and earn up to 20% in referral rewards — invite more, earn more!</td> </tr> <tr> <td width="180"><a href="https://runapi.host/register?aff=iOKB"><img src="assets/partners/logos/runapi.jpg" alt="RunAPI" width="150"></a></td> <td>Thanks to RunAPI for sponsoring this project! RunAPI is a high-performance and reliable AI model API gateway — one API key gives you access to 150+ mainstream models including OpenAI, Claude, Gemini, DeepSeek, and Grok, with prices as low as 10% of the official rate and excellent stability. It works seamlessly with Claude Code, OpenClaw, and other tools. Exclusive benefit for CC Switch users: register via <a href="https://runapi.host/register?aff=iOKB">this link</a> and enjoy a 10% discount on your first top-up!</td> </tr> <tr> <td width="180"><a href="https://www.shengsuanyun.com/?from=CH_4HHXMRYF"><img src="assets/partners/logos/shengsuanyun.png" alt="Shengsuanyun" width="150"></a></td> <td>Thanks to Shengsuanyun for sponsoring this project! Shengsuanyun is a super factory serving AI Native Teams — an industrial-grade AI task parallel execution platform. Its model marketplace aggregates Claude, ChatGPT, Gemini, and other domestic and international LLM and multimedia model capabilities with direct supply. Absolutely no reverse engineering or dilution — platform-wide model SLA availability reaches 99.7%, with <a href="https://watch.shengsuanyun.com/status/shengsuanyun">monitoring dashboards</a> showing green across the board. It also offers enterprise-grade custom gateways for fine-grained team cost and permission management, smart routing, security protection, and BYOK (Bring Your Own Key) hosting. The platform charges on a pay-per-use and tokens plan (coming soon) basis, with invoicing available. Register via <a href="https://www.shengsuanyun.com/?from=CH_4HHXMRYF">this link</a> as a new user to receive ¥10 in credits plus a 10% bonus on your first top-up.</td> </tr> <tr> <td width="180"><a href="https://aigocode.app/invite/CC-SWITCH"><img src="assets/partners/logos/aigocode.png" alt="AIGoCode" width="150"></a></td> <td>Thanks to AIGoCode for sponsoring this project! AIGoCode is an all-in-one platform that integrates Claude Code, Codex, and the latest Gemini models, providing you with stable, efficient, and highly cost-effective AI coding services. The platform offers flexible subscription plans, zero risk of account suspension, direct access with no VPN required, and lightning-fast responses. AIGoCode has prepared a special benefit for CC Switch users: if you register via <a href="https://aigocode.app/invite/CC-SWITCH">this link</a>, you'll receive an extra 10% bonus credit on your first top-up!</td> </tr> <tr> <td width="180"><a href="https://aicoding.inc/i/CCSWITCH"><img src="assets/partners/logos/aicoding.jpg" alt="AICoding" width="150"></a></td> <td>Thanks to AICoding for sponsoring this project! AICoding — Global AI Model API Relay Service at Unbeatable Prices! Claude Code at 19% of original price, GPT at just 1%! Trusted by hundreds of enterprises for cost-effective AI services. Supports Claude Code, GPT, Gemini and major domestic models, with enterprise-grade high concurrency, fast invoicing, and 24/7 dedicated technical support. CC Switch users who register via <a href="https://aicoding.inc/i/CCSWITCH">this link</a> get 10% off their first top-up!</td> </tr> <tr> <td width="180"><a href="https://subrouter.ai/register?aff=l3ri"><img src="assets/partners/logos/subrouter-banner.png" alt="SubRouter" width="150"></a></td> <td>Thanks to SubRouter for sponsoring this project! SubRouter is a marketplace and smart routing platform for AI service operators. Merchants can launch operating sites, publish packages, manage users, models, and pricing, while users discover services and access reliable AI models through one unified API. Register via <a href="https://subrouter.ai/register?aff=l3ri">this link</a>!</td> </tr> <tr> <td width="180"><a href="https://agentmarket.fluxapay.xyz/marketplace/tokenplans"><img src="assets/partners/logos/fluxa-banner.png" alt="FluxA TokenPlan" width="150"></a></td> <td>Thanks to FluxA & Baidu AI Cloud for sponsoring this project! FluxA and Baidu AI Cloud jointly launched AgenticPlan, which gives AI agents the ability to buy, manage and use models, APIs and tools on their own. It includes Baidu Qianfan TokenPlan at up to 40% off, with access to flagship models such as DeepSeek V4, GLM 5.2 and Kimi, plus bonus FluxA AgentMarket API credits that unlock 13,000+ paid APIs for search, data scraping, social media, finance, crypto, image generation, video and more.<br>With the user's authorization, an AI agent can also pay with the official Visa card to procure resources on its own, manage API keys, monitor usage and plan renewals — taking agents from "completing tasks autonomously" to genuinely "planning their own budget and completing tasks". Buy via <a href="https://agentmarket.fluxapay.xyz/marketplace/tokenplans">this link</a>!</td> </tr> <tr> <td width="180"><a href="https://88api.ai/sign-up?aff=HSGY"><img src="assets/partners/logos/88api-banner-en.jpg" alt="88API" width="150"></a></td> <td>Thanks to 88API Token Aggregation Platform for sponsoring this project! 88API is a one-stop multi-model API platform operated by a Hong Kong company, built mainly for developers, creators and AI application users. A unified interface gives access to text, image, speech and video models, covering common workflows such as AI coding, smart translation, content creation, voice-over, image generation and video generation. You can add it as a provider and switch to it right inside CC Switch. 88API supports major international payment methods, can issue invoices, and provides enterprise-grade stable service. Register via <a href="https://88api.ai/sign-up?aff=HSGY">this link</a> to receive exclusive bonus credits!</td> </tr> <tr> <td width="180"><a href="https://apikey.fan/register?aff=CCSwitch"><img src="assets/partners/logos/apikey_banner.png" alt="APIKEY.FUN" width="150"></a></td> <td>Thanks to APIKEY.FUN for sponsoring this project! APIKEY.FUN is a professional enterprise-grade AI relay platform dedicated to providing stable, efficient, and low-cost AI model API access for enterprises and individual developers. The platform supports popular mainstream models such as Claude, OpenAI, and Gemini, with prices as low as 7% of official rates. Register through this project's <a href="https://apikey.fan/register?aff=CCSwitch">exclusive link</a> to enjoy an exclusive offer of up to <strong>permanent 5% off top-ups</strong>.</td> </tr> <tr> <td width="180"><a href="https://9527.codes/register?aff=e5zI"><img src="assets/partners/logos/9527-banner.png" alt="9527CODE" width="150"></a></td> <td>Thanks to 9527CODE for sponsoring this project! 9527 CODE is an enterprise-grade, full-capability AI relay platform focused on stability and service quality, providing fast, stable, and high-quality relay services for Claude Code, Codex, and other mainstream AI models. It places greater emphasis on long-term stability, route quality, and sustained availability, offering enterprises and heavy AI users a stable and efficient one-stop model access solution. With one year of continuous operation, 99.9% service stability, and 7×24 human technical support. Register through this project's <a href="https://9527.codes/register?aff=e5zI">exclusive link</a>; new users can contact customer service to receive trial credit, and referral cashback is unlimited!</td> </tr> <tr> <td width="180"><a href="https://console.apito.ai/agent/register/pQBql2buaqiX3dDS"><img src="assets/partners/logos/claudeapi.png" alt="ClaudeAPI" width="150"></a></td> <td>This project is sponsored by <a href="https://console.apito.ai/agent/register/pQBql2buaqiX3dDS">Claude API</a>. Direct Claude API access — connect Claude Code and Agent apps in 3 minutes. New users can claim a free trial credit. Powered by official Anthropic API keys + AWS Bedrock official channels. No reverse engineering, no model degradation. Full support for Opus / Sonnet / Haiku model lineup, with official capabilities preserved including Tool Use, 1M context window, and more. Built for Claude Code power users, Agent engineers, and enterprise engineering teams. Invoicing and dedicated team support available. Click <a href="https://console.apito.ai/agent/register/pQBql2buaqiX3dDS">here</a> to register!</td> </tr> <tr> <td width="180"><a href="https://code0.ai/agent/register/B2XHxGjGmRvqgznY"><img src="assets/partners/logos/code0.png" alt="code0.ai" width="150"></a></td> <td>Thanks to <a href="https://code0.ai/agent/register/B2XHxGjGmRvqgznY">code0.ai</a> for sponsoring this project! code0.ai is an AI coding service platform built for developers, supporting Claude Code, Codex, Gemini, and other mainstream AI coding capabilities. It helps individual developers and teams use AI Agents more stably and efficiently for coding, debugging, refactoring, and automation workflows. ccswitch users can contact customer support via the <a href="https://code0.ai/agent/register/B2XHxGjGmRvqgznY">code0.ai website</a> to claim test credits and experience a reliable AI coding service.</td> </tr> <tr> <td width="180"><a href="https://teamorouter.cn/?utm_source=cc_switch&utm_medium=referral&utm_campaign=ai_directory"><img src="assets/partners/logos/TeamoRouter-banner.png" alt="TeamoRouter" width="150"></a></td> <td>Thanks to TeamoRouter for sponsoring this project! TeamoRouter is an enterprise-grade Agentic LLM gateway built for developers, AI teams, and businesses. Without requiring any subscriptions, it lets you access Claude Code, Codex, Gemini CLI, OpenAI Codex, and other popular AI agents through a single unified API, while offering API pricing at discounts of up to 90%. Unlike typical API relay services, TeamoRouter aggregates hundreds of official model providers and trusted infrastructure partners, including OpenAI, Anthropic, Vertex, Azure, and AWS bedrock. Every provider is verified for 100% Agent protocol compatibility, cache performance, and request traceability, ensuring stable quality instead of reverse-engineered or diluted endpoints. The platform delivers near-official TTFT, 99.6% SLA, enterprise-scale throughput up to 5,000 QPM, and industry-leading cache hit rates that dramatically reduce token costs for long-running agent workflows. TeamoRouter also offers enterprise features including centralized billing, team management, BYOK, smart routing, usage analytics, dynamic provider optimization, and dedicated support. For an even simpler experience, Teamo Desktop lets you use Claude Code, Codex, Gemini CLI, and other popular AI agents with one-click setup—no API key management or manual gateway configuration required. Register via <a href="https://teamorouter.cn/?utm_source=cc_switch&utm_medium=referral&utm_campaign=ai_directory">this link</a> as a new user to receive 10% off your first top-up.</td> </tr> <tr> <td width="180"><a href="https://ppio.com/activity/ccswitch"><img src="assets/partners/logos/ppio-banner.png" alt="PPIO" width="150"></a></td> <td>Thanks to PPIO for sponsoring this project! PPIO is a leading independent Agentic
Switch API providers in one click and manage MCP, Skills, and Prompts in one place — no more hand-editing JSON / TOML / YAML config files.
[图片:Version](在新窗口打开) [图片:Platform](在新窗口打开) [图片:Built with Tauri](在新窗口打开) [图片:Downloads](在新窗口打开)
English | 中文(在新窗口打开) | 日本語(在新窗口打开) | Deutsch(在新窗口打开) | Changelog(在新窗口打开)
Download(在新窗口打开) · Quick Start(在新窗口打开) · Features(在新窗口打开) · FAQ(在新窗口打开) · User Manual(在新窗口打开)
Kimi K3 is Moonshot AI's most capable model and the world's first open 3T-class model. With 2.8 trillion parameters, native vision, and a 1-million-token context window, K3 delivers frontier performance across long-horizon coding, knowledge work, and reasoning. CC Switch makes it easy to configure and switch to Kimi across agentic tools.
Try a Kimi Code plan (中文站(在新窗口打开) | Global(在新窗口打开)), or use the API through the Kimi Open Platform (中文站(在新窗口打开) | Global(在新窗口打开)).
New user top-up bonus: register and complete your first top-up via the API links above to receive 10% of the amount as bonus API credit, up to CNY ¥1,000.
这一版以修复为主,另有一项新接入:MiniMax Code 成为第十个受管应用,供应商、MCP、Skills、提示词、会话浏览与用量统计一并接入,默认模型、登录和会话删除仍由 MiniMax Code 自己管。修复主要集中在经本地代理使用 Codex 的场景,四个会让 Codex 用不下去的问题在本版关闭:Codex 0.154 起放进
additional_tools的工具让严格的 Chat 网关直接 400;一张detail: "original"的截图让整段会话此后每轮都 400;编辑或切走部分 Codex 供应商会悄悄清空已保存的 API Key;绑定到已删除 ChatGPT 账号的卡片既切不走,也改不了绑定。Claude Code 这边,GitHub Copilot 下的 auto mode 恢复可用,/effort的xhigh、max与 Grok 4.6 的推理强度都按所选档位发送。Claude Desktop 的第三方配置现在也支持 Linux。本版含数据库迁移(v18 → v19),建议升级前手动备份一次数据库,见升级提醒(在新窗口打开)。
English →(在新窗口打开) | 日本語版 →(在新窗口打开)
~/.minimax/config.yaml,与 MiniMax Code 自己的设置共存,多个供应商可以同时存在;MCP 与 ~/.minimax/mcp.json 双向同步;Skills 与 AGENTS.md 提示词库一并接入;预设来自 Pi 的预设目录,共 41 个。默认模型、登录与会话删除仍由 MiniMax Code 自己管,CC Switch 不碰。additional_tools 里的工具正确并入请求,不再变成一条 content: null 的 system 消息。/effort 选哪档就发哪档(#7318(在新窗口打开)、#7531(在新窗口打开),修复 #7314(在新窗口打开)、#5367(在新窗口打开)):xhigh 不再被丢,映射到 Grok 4.6 的模型重新带上推理强度,GPT-5.6 与 GPT-6 Astra 的 max 不再降成 xhigh。[!WARNING]
唯一官方渠道声明(请务必阅读)
CC Switch 是完全免费、开源的桌面应用,不会向用户收取任何费用。请仅通过下列官方渠道获取本软件:
类别 唯一官方 官网 ccswitch.io(在新窗口打开) 源码 github.com/farion1231/cc-switch(在新窗口打开) 下载 GitHub Releases(在新窗口打开) 作者 @farion1231(在新窗口打开) 举报山寨 GitHub Issues(在新窗口打开) 任何向你收费、要求充值、或索取登录凭据的"CC Switch"网站或客户端均为假冒。如果你被诱导支付了费用,请立即停止操作并通过 GitHub Issues 反馈。
MiniMax Code 是 MiniMax 的编程智能体,终端版 mcode 与桌面版共用 ~/.minimax 数据目录。CC Switch 以累加模式管理它,与 OpenCode、Hermes、Pi 相同:每个供应商是 ~/.minimax/config.yaml 中 custom_provider 下的一项,多个供应商共存,写入时保留 MiniMax Code 自己的其他设置,并遵守它的配置文件锁。MCP 与 ~/.minimax/mcp.json 双向同步,Skills 部署到 ~/.minimax/skills/,提示词库写入 ~/.minimax/AGENTS.md;会话浏览是只读的,用量从 MiniMax Code 自己的数据库增量导入。接入范围刻意收得很紧:默认模型、登录、云端功能与会话删除都留给 MiniMax Code,本地代理接管、故障转移、托盘、通用配置片段与 Profiles 也不覆盖它。(#7383(在新窗口打开)、#7578(在新窗口打开))
修复里分量最重的是经本地代理使用 Codex 的一组。Codex 0.154 起把插件、执行命名空间等额外工具放进一个 additional_tools 输入项,转换器把它当成消息,生成一条 content: null 的 system 消息,严格的 Chat 网关直接 400,里面的工具也被丢掉(#7454(在新窗口打开))。Codex 给截图标 detail: "original",这个值只在 Responses API 里合法,被原样搬进 Chat 请求后,方舟/豆包、日日新、GLM、vLLM 一类网关对整条请求返回 400;Codex 每轮都会重放完整历史,一张图就让这个会话从此每轮失败(#7476(在新窗口打开))。另外两个是账号数据问题:自带凭据来源的 Codex 供应商在编辑保存或切走时,已保存的 API Key 被 live 里的空 auth 覆盖(#7434(在新窗口打开));在认证中心删掉并重新登录 ChatGPT 账号后,绑着旧账号的卡片每次切换、接管和启动恢复都报「账号不存在」(#7395(在新窗口打开))。
Claude Code 经转换的请求也有几处修正:没写描述的工具不再以 "description": null 发出(#7319(在新窗口打开)、#7378(在新窗口打开));GitHub Copilot 下 auto mode 的分类器不再因 stop 参数 400(#5404(在新窗口打开));/effort xhigh 不再被丢,映射到 Grok 4.6 的模型重新带上推理强度,GPT-5.6 与 GPT-6 Astra 的 max 不再降成 xhigh(#7318(在新窗口打开)、#7531(在新窗口打开))。其余修复覆盖 Skills 安装(skills.sh 的 id 与目录名不一致、超过 1 万个文件的仓库,#6381(在新窗口打开)、#7489(在新窗口打开))、提示词读回外部修改(#7194(在新窗口打开))、WSL 上的 OMO 统一配置(#7550(在新窗口打开))、关于页的版本探测(#7346(在新窗口打开)、#7348(在新窗口打开))与 Windows 任务栏(#6348(在新窗口打开))。新功能方面,Claude Desktop 第三方配置支持 Linux(#7331(在新窗口打开)),请求日志显示每秒输出 token 数(#3369(在新窗口打开)),OpenCode 供应商表单可以搜索并批量添加获取到的模型(#7515(在新窗口打开))。
预设做了一轮维护:国内 Codex 预设按各家 Responses 文档对齐(Kimi For Coding 升到 K2.8 Preview、1M 窗口,DeepSeek 默认 deepseek-flash),MiniMax 国内站与百灵迁到当前官方域名,AICodeWith 的 Codex 族端点改正为 /v1,另新增几个预设。DeepSeek 在 2026-09-12 撤回了 V4 Pro 改按 V4.1 Flash 计价的公告,v3.20.3 提前执行的调价在本版改回高峰档。本版含数据库迁移:SCHEMA_VERSION 18 → 19,为 MCP 与 Skills 表加上 MiniMax Code 的启用列。
发布日期:2026-09-22
更新规模:46 commits | 156 files changed | +8,673 / -684 lines
MiniMax Code(app_type 为 mcode)的终端版与桌面版共用一个数据目录,默认是 ~/.minimax,可以用 MINIMAX_DATA_DIR(或旧的 MAVIS_DATA_DIR)改到别处。(#7383(在新窗口打开)、#7578(在新窗口打开))
~/.minimax/config.yaml 中 custom_provider 下的一项,卡片算不算「已添加」只看文件里有没有它的键。表单沿用 OpenCode 风格的结构化编辑器,另有 JSON 编辑器,支持 anthropic-messages、openai-completions、openai-responses 三种 API 格式;预设从 Pi 的预设目录派生,发版时共 41 个。在 MiniMax Code 里新建的自定义供应商会在列表加载时导入,在那边做的修改也会同步回来;账号类供应商(kind 不是 custom 的)既不导入也不覆盖。写入时保留 config.yaml 里其他所有键,遵守 MiniMax Code 自己的 config.yaml.lock 目录锁(超过 10 秒的锁会被回收,否则保存失败并提示重试),原生文件写入失败时数据库改动回滚。在 config.yaml 里删掉某个供应商后,它的卡片会立即刷新为未添加,可以再次添加(#7578(在新窗口打开))。defaultModel / defaultLightModel。某个供应商或模型正被这两项引用时,CC Switch 拒绝删除或停用它,需要先在 MiniMax Code 里换一个默认模型。~/.minimax/mcp.json 双向同步。编辑只替换传输字段(command、args、env、url、headers、type),timeout 这类原生选项会保留;同 id 但传输方式不同的服务器会跳过并提示,不会覆盖;全量重新同步不会删除开关关闭的条目,只有显式停用或删除才会移除。数据库提交失败时原生文件会还原。~/.minimax/skills/,沿用 Pi 的归属检查,同名但不是 CC Switch 部署的 Skill 一律不动。提示词库把启用的提示词写进 ~/.minimax/AGENTS.md:切换前先把文件里的外部修改写回当前提示词;还没有启用任何 MiniMax Code 提示词时,文件里对不上的内容先存成一条备份提示词。停用最后一个提示词时清空文件;单个提示词上限 32 KiB(按 UTF-8 字节计)。apps 支持 mcode;MiniMax Code 的供应商 deep link 会被拒绝。会话浏览器以只读方式打开 ~/.minimax/v2/sqlite/runtime-state.sqlite,列出 MiniMax Code 的对话(隐藏、归档、子会话以及 peek/channel/cron 会话不列),每个会话提供 resume 命令 env MINIMAX_DATA_DIR=<dir> mcode --session <id>(Windows 上是 PowerShell 写法)。会话与 MiniMax Code 的运行状态绑定,所以界面上不能删除,后端也会拒绝。
用量从同一个数据库的 local_runtime_token_usage 表增量导入,来源为「MiniMax Code (Session)」,看板有独立的应用筛选。推理 token 计入输出;模型 id 取第一个 / 之后的部分(custom_provider:router/vendor/model 记为 vendor/model)。MiniMax Code 自己算出的 cost_usd 是有限的非负数时直接采用,缺失、为负或非有限值时改按 CC Switch 的定价表计算;原生费用恰好为 $0 的行按 $0 入库,但和其他有 token 的 $0 行一样,一旦该模型在定价表里有了价格,就会被共享的费用回填重新计价。与 Pi 一样,含 MiniMax Code 的合计带「缓存写入数据不完整」的提示。(#7383(在新窗口打开))
此前在 Linux 上写 Claude Desktop 第三方配置会直
如今主流开源模型的官方 API 大多已原生支持 OpenAI Responses 格式。本版把 Codex 里 Kimi 的两条预设也从 Chat Completions 转换改为原生 Responses 直连——至此 DeepSeek、智谱 GLM、千问、MiniMax、小米 MiMo、LongCat、Kimi 这些主流开源模型的官方 Codex 预设全部直连厂商端点,连同火山豆包与腾讯混元,不再需要开本地路由做格式转换。如果你的 Codex 卡片还是早先添加的 Chat 格式,重新添加一次预设,或在编辑页把「上游格式」改成 Responses,就能直连。其余是一波由贡献者主导的正确性修复:空
reasoning_content占位符不再让 Claude Code 被空 Thought 块刷屏、Codex 长任务不再在一句进度汇报后戛然而止、Claude Desktop 的模型探针不再在 Responses 上游下误报「不可用」、每次退出都把 Claude 的代理重试与超时抄给 Codex/Gemini/Grok Build 的串写关闭、统一供应商同步不再清空子卡设置;Codex 侧补上省略model_provider时的代理路由、Windows 上正在增长的会话用量、托盘里托管账号的额度。预设与定价做了一轮维护:聚合平台目录刷新、千问AI平台改名并升到 Qwen 3.8、MiniMax 默认 M3、DeepSeek V4 家族按 V4.1 Flash 档调价。本版不含数据库迁移。
English →(在新窗口打开) | 日本語版 →(在新窗口打开)
openai_chat 改为 openai_responses,Codex 直接连厂商的 /v1/responses,不再需要开本地路由做 Responses→Chat 转换。至此 DeepSeek、智谱 GLM、千问、MiniMax、小米 MiMo、LongCat、Kimi 这些主流开源模型的官方 Codex 预设全部直连;仍只提供 Chat Completions 的端点(百度千帆、腾讯 Token Plan、QwenCloud For Coding、StepFun、百灵、ModelScope 与各聚合平台)继续经本地路由转换。如果你现有的卡片还是早先添加的 Chat 格式(Kimi,以及前几版已切换的 DeepSeek、GLM 等),重新添加一次预设,或在编辑页把「上游格式」改成 Responses,即可直连;详见升级提醒。reasoning_content 占位的上游,不再让每个 token 变成一行外加一个空思考块。stop。max_tokens=1 在转换时夹到 Responses API 允许的最小值 16。model_provider 的 Codex 卡接管时也走本地代理(#7263(在新窗口打开),修复 #6256(在新窗口打开)),不再悄悄直连 api.openai.com。limits[] 数组。deepseek-flash 与 V4 家族按 V4.1 Flash 档记账,不再按 $0 或旧高峰价。model_provider 时的路由修复落在这条路径。[!WARNING]
唯一官方渠道声明(请务必阅读)
CC Switch 是完全免费、开源的桌面应用,不会向用户收取任何费用。请仅通过下列官方渠道获取本软件:
类别 唯一官方 官网 ccswitch.io(在新窗口打开) 源码 github.com/farion1231/cc-switch(在新窗口打开) 下载 GitHub Releases(在新窗口打开) 作者 @farion1231(在新窗口打开) 举报山寨 GitHub Issues(在新窗口打开) 任何向你收费、要求充值、或索取登录凭据的"CC Switch"网站或客户端均为假冒。如果你被诱导支付了费用,请立即停止操作并通过 GitHub Issues 反馈。
主流开源模型的官方 API 如今大多已原生提供 OpenAI Responses 端点,Codex 也早已以 Responses 为唯一原生协议。此前 CC Switch 对这些模型的 Codex 预设分两种形态:厂商端点原生支持 Responses 的直连,只提供 Chat Completions 的则由本地代理把 Codex 的 Responses 请求转换成 Chat、再把流式响应转换回来。转换路径能用,但多一层翻译就多一处出错的地方——本版修的 #7280(在新窗口打开) 正是这类转换缺陷。Kimi 开放平台与 Kimi For Coding 的端点现在都原生提供 /v1/responses,官方 Codex 接入指南也要求 wire_api = "responses",于是这两条预设改为原生 Responses 直连。至此 DeepSeek、智谱 GLM、千问(千问AI平台 / QwenCloud)、MiniMax、小米 MiMo、LongCat、Kimi 这些主流开源模型的官方 Codex 预设全部直连厂商端点,连同火山豆包与腾讯混元;仍只提供 Chat Completions 的端点(百度千帆、腾讯 Token Plan、QwenCloud For Coding、StepFun、百灵、ModelScope 与各聚合平台)继续经本地路由转换,手册的「仅 Chat」示例也改为点名这些供应商。
其余是一波由贡献者主导的正确性修复,关闭了几个挂了数月的 issue。代理侧:在每个 chunk 里保留空 reasoning_content 占位的 OpenAI 兼容上游不再让 Claude Code 被空 Thought 块刷屏、逐 token 换行;Codex Responses→Chat 转换器不再把 commentary 消息与紧随的工具调用拆成两条 assistant 消息——这曾让长任务在一句进度汇报后就结束;Claude Desktop 的一 token 模型探针夹到 Responses API 的最小值,映射模型不再被报为「不可用」;Codex 路由下的图片生成补上贴完整端点、大小写后缀与流式用量三处。两个数据完整性问题关闭:每次正常退出都把 Claude 的重试与超时设置抄到 Codex、Gemini、Grok Build 的代理行;同步统一供应商会清空子卡的用量脚本、通用配置退出与端点自动选择,并把卡片挤到列表底部。Codex 侧:卡片省略 model_provider 时接管现在尊重代理地址;用量导入经持久化的字节游标在 Windows NTFS 上识别正在增长的 rollout;托盘为托管 Codex 卡显示所绑 ChatGPT 账号的额度;Claude Fable 的周限额出现在卡片与托盘。Claude 供应商编辑器新增「禁用 Artifact 工具」快捷开关,给拒收 Claude Code Artifact 工具 schema 的网关用。
预设侧:聚合平台 Codex 预设刷新到当前目录,DashScope/百炼改名千问AI平台并升到 Qwen 3.8(#7183(在新窗口打开)),MiniMax 默认改到 M3(#7255(在新窗口打开)),内置 DeepSeek Codex 目录镜像支持视觉的 deepseek-flash(#7286(在新窗口打开)),DeepSeek V4 家族按 V4.1 Flash 档调价。本版不改数据库 schema。
发布日期:2026-09-11
更新规模:22 commits | 62 files changed | +3,458 / -802 lines
部分第三方 Anthropic 兼容网关(DeepSeek 在列)用严格的正则校验器检查工具 JSON Schema,拒收 Claude Code 的 Artifact 工具在其 str_replace 数据库能力铺开后发出的 Unicode 属性转义(\p{Cc}、\p{Cf} 等)——此后每个请求都 400 Invalid schema for function 'Artifact',与所选模型无关。Claude 快捷开关旁新增第六个复选框,在供应商上设置 env.CLAUDE_CODE_DISABLE_ARTIFACT="1",把 Artifact 工具整个排除出 tools 数组;取消勾选时像其他开关一样删掉该键。四语文案与手册(zh/en/ja)的开关表已同步。
Claude OAuth 用量 API 现在把按模型的周限额放在顶层 limits[] 数组里(kind: "weekly_scoped",带 scope.model.display_name 与 percent),不再是独立的顶层窗口;解析器只读旧窗口,Fable 的限额从未显示过。limits[] 现在解析进 seven_day_fable、seven_day_opus 与 seven_day_sonnet——scoped 行覆盖同名旧窗口,重复与畸形行跳过,旧窗口、额外用量与未知窗口都保留。托盘给 Fable 单独一组标签,不再并进周 max 值;四语补「Fable」标签。
千问AI平台 Token Plan(#7183(在新窗口打开))加入 Claude Code、Claude Desktop、Codex、Hermes、OpenClaw、OpenCode 与 Pi 七个应用的预设库,与下文的改名同批。存量供应商不受影响。
deepseek-flash 与 deepseek-v4-flash-vision-exp 的定价行deepseek-flash 是 DeepSeek 目前唯一推荐的 id,deepseek-v4-flash-vision-exp 是官方安装脚本 1.2.0 之前写入的旧视觉名;两者都由 V4.1 Flash 服务,按其 $0.30/$1.20 每百万、缓存读 $0.006 计费。此前两者都没有行,而 LIKE '{id}-%' 的前缀回退只匹配更长的 id,这两个 id 下的请求一律按 $0 入账。
Kimi 开放平台(api.moonshot.cn/v1)与 Kimi For Coding(api.kimi.com/coding/v1)现在都原生提供 /v1/responses,官方 Codex 接入指南要求 wire_api = "responses",于是两条预设从 openai_chat(本地代理做 Responses→Chat 转换、必须开路由接管)改为 openai_responses(Codex 直连)。两个端点都用真实密钥按 Codex 0.153.4 的完整请求形态实测(reasoning.encrypted_content、reasoning.summary、托管 web_search、回放的 reasoning 项),并用 codex exec 跑通端到端工具循环。kimi-k3 成为开放平台的默认模型与目录首行,kimi-k2.7-code 保留为第二行;只对 Chat 转换有意义的 codexChatReasoning 从两条预设移除,Kimi For Coding 的 promptCacheRouting 也去掉——原生路径上 Codex 自己发 prompt_cache_key;每个 Kimi 行声明 supportsParallelToolCalls,与官方 Kimi Code models.json 一致;不声明 reasoning 默认值,因为 config.toml 的 model_reasoning_effort 优先于目录默认值,后者只标记 /model 选择器。手册(zh/en/ja)的「仅 Chat」示例改为点名仍走 Chat 路由的供应商,「Kimi Code 拒收 codex-cli user agent」的过时说明已纠正。无后端改动。
SiliconFlow(.cn)默认改为 deepseek-ai/DeepSeek-V4-Flash(1M 窗口,high/max 档,显式的 enable_thinking/reasoning_effort 约定),因为国内站的 MiniMax M2.5 已下线;Atlas Cloud 改为 zai-org/glm-5.2,其 Coding Plan 包含的最新 GLM(1M 窗口);Novita 改为 zai-org/glm-5.3(1M,纯文本);NVIDIA NIM 改为 moonshotai/kimi-k3(1M,文本+图片;NIM 接受 reasoning_effort 的 low/high/max、字段缺省时按 max,预设显式钉 high 与其 config.toml 一致);OpenCode Go 的 GLM/Kimi 条目更新——glm-5.3(新默认)、glm-5.3-flash 与 kimi-k3 替换 glm-5.2、glm-5.1 与 kimi-k2.7-code,窗口、模态与档位按 models.dev 镜像,DeepSeek V4 Pro/Flash 与 MiMo V2.5 Pro 行保留。平台未文档化新模型思考/档位约定的(Atlas Cloud、Novita),预设带一条显式的空覆盖,后端不再按模型名注入厂商原生思考字段。
国内百炼预设改名千问AI平台,配专属图标,控制台与 API Key 链接改到 platform.qianwenai.com;Qwen 系列在 Claude Code、Claude Desktop、Codex、Hermes、OpenClaw、OpenCode 与 Pi 七处升到 3.8 代——Opus/Sonnet/Haiku 角色对应 qwen3.8-max / qwen3.7-plus / qwen3.8-flash,统一使用官方的 983,616 token 窗口;国内 Bailian For Coding 预设(Claude Code、Claude Desktop、Hermes)改名 千问AI平台 Coding Plan,端点不变;OpenClaw/OpenCode/Pi 的模型元数据对齐官方客户端文档。国际 QwenCloud 预设保留名称、换专属图标;按量付费预设升到 Qwen 3.8,Token Plan 预设与之对齐(Codex 与 Hermes 本已带 3.8 行),QwenCloud For Cod
这一版仍以 Codex 为主线,但形态从上一版的「一次重设计」变成「一波兼容性修复」:Grok 终于能经 xAI 原生 Responses API 在 Codex 路由下跑通——xAI 拒收的工具 schema、Codex 拒收的整数浮点、多智能体注入的邮箱消息、xAI 不认识的 Codex 角色型号,四道坎逐一拆掉;围绕它的一族「一因一果」修复:Grok OAuth 卡不再被 v3.20.1 的切换闸误拦、接管不再把 Codex 卡在登录屏、GPT-6 走 Codex OAuth 不再报「需要更新 Codex」、Claude Code 在 Codex OAuth 上恢复并行工具调用、内置图片生成走通本地路由、四个 catalog 缺陷(旧版 Codex 拒载、DeepSeek 下 MCP 工具全隐身、视觉型号被判纯文本、Kimi 工具 schema 400)关闭,智谱 GLM 预设改指官方 Responses 端点。用量侧,Codex resume 后统计停摆与代理每轮打断前缀缓存两条硬修复落地。预设阵营新增腾讯云 Token Plan、QwenCloud、AICodeWith 与两家赞助商 9527CODE、SoleAPI,定价表补入七个新模型并做了一轮九月调价。本版不含数据库迁移。
English →(在新窗口打开) | 日本語版 →(在新窗口打开)
agent_message 被翻译成普通消息、gpt-5.6-sol 一类角色型号被映射到你配置的 Grok 模型——子智能体不再 422,工具调用不再被拒。auth.json 之后不再把 Codex 困在登录屏。parallel_tool_calls 默认改为 true,每轮不再只允许一个工具调用。/images/generations 与 /images/edits 两条路由经本地代理透传,用量按 token 记账。tool_search,MCP 工具直接列出。$ref 只对 Moonshot 域名改写成 allOf,其他供应商的 schema 逐字节不变。/api/v1、默认 glm-5.3,存量卡需重新导入。[!WARNING]
唯一官方渠道声明(请务必阅读)
CC Switch 是完全免费、开源的桌面应用,不会向用户收取任何费用。请仅通过下列官方渠道获取本软件:
类别 唯一官方 官网 ccswitch.io(在新窗口打开) 源码 github.com/farion1231/cc-switch(在新窗口打开) 下载 GitHub Releases(在新窗口打开) 作者 @farion1231(在新窗口打开) 举报山寨 GitHub Issues(在新窗口打开) 任何向你收费、要求充值、或索取登录凭据的"CC Switch"网站或客户端均为假冒。如果你被诱导支付了费用,请立即停止操作并通过 GitHub Issues 反馈。
v3.20.1 之后的开发依旧由 Codex 领衔,但这一次是一波兼容性修复而不是一次重设计。最长的一条线是 Grok:xAI 的原生 Responses API 在几处与 Codex 的假设相左——它在采样前就拒收 Codex Desktop 的工具 schema、Grok 回传的 JSON 整数带小数点而 Codex 的解析器拒收、Codex 多智能体模式注入的 agent_message 邮箱项 xAI 无法反序列化、Codex 自己的角色型号(如 gpt-5.6-sol)xAI 直接 404。本版把这四道坎在同一个原生 Responses 闸门后逐一拆掉,Grok 终于能在 Codex 路由下完整工作,包括子智能体。
围绕它是一族「一因一果」的修复:v3.20.1 引入的无密钥安全闸误拦了代理注入令牌的 Grok OAuth 卡;直接切换删掉 auth.json 之后再开接管,Codex 会被过时的 requires_openai_auth = true 困在登录屏;GPT-6 走 Codex OAuth 接管被后端以客户端版本过旧拒绝;Claude Code 在 Codex OAuth 上被强制串行调用工具;内置图片生成在本地路由下 404;四个 catalog 缺陷——旧版 Codex 拒载 catalog、DeepSeek 预设下 MCP 工具全部隐身、DeepSeek 视觉型号被判纯文本、Kimi 工具 schema 必现 400——各自关闭;智谱 GLM 预设从 Chat 端点改指官方 Responses 端点。用量侧两条硬修复:Codex resume 之后统计静默停摆,以及代理把 Claude Code 每轮注入的中途 system 消息合并到开头、打断上游前缀缓存。
预设库新增腾讯云 Token Plan、QwenCloud、AICodeWith 与赞助商 9527CODE、SoleAPI,Pi 补齐腾讯与 PPIO;定价表补入七个新模型行并按各厂商九月官方价页做了一轮调价。本版不改数据库 schema。
发布日期:2026-09-07
更新规模:52 commits | 71 files changed | +10,483 / -573 lines
腾讯云 Token Plan(六产品 × 六应用,#7011(在新窗口打开))、QwenCloud(三计划 × 七应用,#6214(在新窗口打开))、AICodeWith(八应用)与两家赞助商 9527CODE、SoleAPI(九应用)加入预设库;Pi 补齐腾讯 TokenHub / Token Plan(#7159(在新窗口打开))与 PPIO(#6870(在新窗口打开))。存量供应商不受影响;需要重新导入的情形见升级提醒。
Claude Fable 5.1 与 Mythos 5.1:$10/$50 每百万,缓存读 $0.25、缓存写 $12.50(#7051(在新窗口打开),修复 #7050(在新窗口打开))——此前没有行,这些请求一律按 $0 入账,因为前缀规则只找更长的带日期 id、无法回落到 claude-fable-5;GLM-5.3:$1.40/$4.40,缓存读 $0.26,与 Z.ai 官方价一致(#6591(在新窗口打开));GLM-5.3 Flash:$0.15/$0.50,缓存读 $0.03(#7163(在新窗口打开));GPT-6 Astra:$10/$50,缓存读 $1、缓存写 $12.50,low/medium/high/xhigh 后缀回落到基础行(#7162(在新窗口打开));Gemini 3.8 Flash:$0.75/$3.75,缓存读 $0.075,无缓存写费(#7164(在新窗口打开));Qwen3.8 Flash:$0.15/$0.47,1M 窗口内不分档,缓存读/写 $0.016/$0.20。全部只是 seed 行、不改 schema;存量安装下次启动即拿到,启动时的成本回填会把这些 id 此前按 $0 记录的历史行补上价格。
Sonnet 5 改回 $2/$10——Anthropic 价页现已注明介绍价即标准价、9 月 1 日涨到 $3/$15 的计划不再执行——并带一条守卫修复:仍是 seed 值 3/15/0.30/3.75 的安装被纠正,用户自定义过的行不动(#7051(在新窗口打开))。GPT-5.6 Sol、裸名 gpt-5.6 及其五个档位后缀行同步从 $5/$30(缓存读 $0.50、写 $6.25)降到促销价 $4/$20(缓存读 $0.40、写 $5),至少持续到 2026-11-21。Gemini 3.6 Flash 从 $1.50/$7.50/$0.15 降到与 3.8 Flash 相同的介绍价 $0.75/$3.75/$0.075,至 2026-12-31。MiniMax M2、M2.1、M2.5 统一为官方按量价 $0.30/$1.20,缓存读 $0.03、缓存写 $0.375——后者此前记为零。每项调价都是 seed 加一条按旧值守卫的修复条目,追加在既有链条之后,老数据库会逐级经过中间价位。
一组贡献者提交关闭了原生 Responses 通往 xAI 路径上四个互相独立的故障。xAI 在采样前就拒收 Codex Desktop 的内置工具 schema,于是请求路径上折叠根级 oneOf/anyOf 函数参数(#6815(在新窗口打开)),各分支的 required 取交集而不是并集,压平后的 oneOf 不会强求所选分支没有的字段。Grok 回传的 JSON 整数带小数点,Codex 的解析器拒收为整数,于是已完成 function_call 参数里的整数浮点被改写(恰好 2^64 现在直接拒绝而不是静默改写差一)。Codex 多智能体模式注入的 agent_message 邮箱项 xAI 无法反序列化,每个子任务在任何工具运行前就 422,于是这些项被改写成普通 user 消息、加密内容压平为文本。xAI 对 Codex 的角色型号(如 gpt-5.6-sol)直接 404,于是未知的请求模型被映射到供应商配置的模型(catalog 里的 grok-4.5 一类与任何 grok 前缀 id 原样放行),且映射先于清洗执行,落到 grok-4.5 上的子智能体也会被剥掉不支持的采样字段。全部逻辑收在一个原生 Responses 闸门后的独立模块里,便于在上游覆盖同样场景之前 rebase 或 cherry-pick。grok-4.5 的 xAI 预设现在声明 low/medium/high/xhigh:2026-08-30 实测该端点接受这四档、对 max 报 HTTP 400,而 Codex 不会把目录外的档位夹回范围。(#6917(在新窗口打开))
v3.20.1 的 config-only 切换重构把无密钥安全闸推广到所有写入路径,误伤了代理托管的 OAuth 卡(xAI Grok OAuth):它们本来就无密钥——本地代理逐请求注入真实令牌——但预设快照继承了 0.149 之前模板的 requires_openai_auth = true,安全闸把它当作「会回落到官方登录」而拒绝切换。修的是快照而不是给闸门开例外:对代理注入的 OAuth 供应商(xAI OAuth 与 GitHub Copilot;Codex OAuth 刻意排除,官方登录就是它的凭据)在活跃自定义表上把该标志强制为 false,挂在共享的有效供应商构建器里,预检、普通写入与代理备份/接管投影看到同一形状,Codex 0.149 无需读 auth.json 即判定为未鉴权。预设源头现在就输出 false,存量卡在下一次切换时自愈。
「非接管切换时保留官方登录」关闭(默认)时,直接切换到第三方供应商会删除 ~/.codex/auth.json;此后再开启代理接管,config.toml 从存量卡重建,而卡上仍带 0.149 之前的 requires_openai_auth = true,于是 Codex ≥ 0.149 即使已有代理占位 bearer token 也停在登录屏——接管中执行 codex logout 后热切换也踩同一个坑。接管写入器现在按 Codex 自己会观察到的登录状态戳记活跃自定义表的标志,与直接切换的计划对齐:先按 Codex 的规则解析鉴权模式(显式 auth_mode 优先,其次个人访问令牌、Bedrock API 密钥、Bedrock 访问密钥、OPENAI_API_KEY、ChatGPT)再核对凭据,紧邻过期 API 密钥的 Bedrock 凭据永远不会被抬举为 OpenAI 登录;先判定凭据存储再碰 auth.json——keyring 与 auto 存储无法从磁盘判定、保留卡上原值,ephemeral 存储一律视为未登录,只有 file 存储才读文件,文件缺失、不可读或损坏都视为未登录且不再让接管写入失败。代理注入的 OAuth 卡无论磁盘上是什么都保持中和后的 false;官方透传与托管官方分支不动。
普通托管账号登录若 ChatGPT workspace 与 稳定用户身份(id_token 的 subject)同时匹配已有账号,会以本地化提示拒绝而不是新建第二行。检查在存储锁内执行,并发完成的登录无法漏进重复;刷新锁现在只在定向重鉴权时获取,被拒的添加不会留下锁条目。同一 workspace 的不同用户仍可并存。(#7061(在新窗口打开))
Claude 转 Codex OAuth 路由上的 gpt-6-astra 请求自报 Codex 0.144.1,低于该模型 0.153.0 的最低客户端版本,ChatGPT 后端回 HTTP 400。自报版本升为 0.153.4,originator 与版本常量在生成与模型发现之间共享——后者此前发的是 cc-switch 自己的包版本与 orig
社区评论
登录后即可分享使用体验,首条评论会先经过审核。前往登录