What is jc-proxy?

This project is a small server that sits between your AI applications and various AI companies, letting you use one consistent connection point no matter which AI provider is actually answering behind the scenes. It understands the same request formats used by OpenAI and Anthropic, so tools built for either of those can talk to it without changes. The author built this after running into rate limits on free tiers of services like Google, OpenRouter, and Groq while experimenting with AI agent tools such as n8n and Hermes. Paid services work fine but cost money, which is a real barrier for students and hobbyists. Existing proxy tools were either built for large companies or too complicated to run alone. This project's main feature is automatic failover: when one free provider hits its rate limit or fails, the gateway automatically switches to a backup provider so an AI agent workflow does not break. It supports several providers including Groq, Gemini, OpenRouter, Cloudflare Workers AI, and anything else that speaks the OpenAI format, such as locally run Ollama or vLLM setups. It also includes a web based admin dashboard for managing provider settings, checking status, and reviewing routing logs, along with support for streaming responses and tool calling, which lets an AI model trigger actions during a conversation. It can be run directly with Node.js or through Docker, and the Docker setup includes an optional local web search tool called SearXNG. The project is specifically designed to work well with Claude Code and n8n AI Agents, but is also compatible with other agent frameworks like LangChain and LlamaIndex. The readme includes a clear warning that since this proxy holds your private API keys, it should never be exposed directly to the public internet without proper authentication in front of it. It is released under the MIT license.