{"repo":"1837620622/cto-new-openai-proxy","free":true,"listed":false,"github":"https://github.com/1837620622/cto-new-openai-proxy","clone":"git clone https://github.com/1837620622/cto-new-openai-proxy.git","description":"🚀 cto.new 免费反代 | OpenAI & Anthropic 兼容 API | GPT-5.4 / GLM-5.1 账号池自动轮询 · JWT 刷新 · Dashboard UI · 批量注册 ｜ Free OpenAI/Anthropic-compatible API proxy with account pool auto-rotation, JWT refresh, bulk registration & real-time dashboard. ⚠️ 不支持 agent 工具调用 / No function calling support.","language":"Python","stars":36,"topics":["account-pool","ai-proxy","anthropic-api","api-proxy","chatbot-api","chatgpt-proxy","claude-api","clerk-auth","free-api","glm"],"license":"MIT","category":"networking-infra","readme_excerpt":"cto.new Free API Proxy · cto.new 免费反代 Batch-register cto.new accounts and expose GPT-5.4 / GLM-5.1 as free OpenAI & Anthropic compatible APIs. 批量注册 cto.new 账号，把 GPT-5.4 / GLM-5.1 免费模型以 OpenAI /v1/chat/completions 和 Anthropic /v1/messages 兼容接口的形式暴露出来。 English · 中文 --- English What it does A Go reverse proxy that turns cto.new (Engine Labs) free chat models into standard OpenAI / Anthropic compatible APIs, with automatic account rotation from a bulk-registered pool. - Models : gpt-5.4 (3× quota), glm-5.1 (2× quota), plus aliases ( gpt-4o , claude-sonnet-4.6 , o1/o3 → all route to gpt 5 4 ) - Protocols : OpenAI chat completions (stream + non-stream), Anthropic messages (stream + non-stream), multimodal input - Account pool : LRU selection, auto JWT refresh via Clerk client cookie, quota/dead marking with 6h cooldown, 30s hot-reload of accounts.jsonl - Dashboard : Web UI at http://127.0.0.1:9091 with tabs for Overview, Accounts, Request Logs, Usage - Bulk registration : Playwright / Selenium scripts to create accounts, plus a login-based cookie refresher for when client expires ⚠️ Important: No agent / function calling support This proxy does not support OpenAI function calling or Anthropic tool use . Reason: cto.new's /engine-agent/chat endpoint is itself a full Engine Labs Coding Agent with its own sandbox and system prompt — we can't hijack tool-call protocol at the proxy layer. Even with OpenAI-standard tools field injection + aggressive prompt engineering, the upstream model","default_branch":null,"files":null,"tree":[],"storefront":"/r/1837620622","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/1837620622/cto-new-openai-proxy/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}