free-one-api

Why we included this project

free-one-api takes a set of reverse-engineered LLM libraries, like the unofficial ChatGPT, Claude, and Qwen adapters, and exposes them through the standard OpenAI request format, so code already written against OpenAI's API keeps working even when you switch to a free or unofficial backend underneath. It also handles the operational overhead: automatic load balancing across channels, heartbeat checks that pull dead endpoints out of rotation, stream mode for chat-style replies, and a small web UI for managing channels and checking logs. Developers exploring free LLM access or putting together budget prototypes get one stable interface while the adapters behind it keep changing. Because those backends are reverse-engineered and naturally unstable, treat it as a tool for evaluation and tinkering rather than a place to park critical production traffic.