copilot-api

Why we included this project

Keeping several coding assistants working usually means managing a separate login and endpoint for each one. This project runs a single local server that accepts OpenAI Chat Completions, the OpenAI Responses API, and Anthropic Messages, so Claude Code, OpenCode, and Codex can all talk to one address instead of carrying their own credentials. From there it routes requests to GitHub Copilot, a built-in Codex provider, or third-party options like DeepSeek, Kimi, and OpenRouter. That is handy when you want to consolidate paid subscriptions, or when you want people to switch providers without rewriting their clients. The server ships ready to run, with a login command and Docker support, so a working gateway can be up in a few minutes and the interesting work is deciding which models to expose. One caveat: it needs Node 22.x or Bun, and token-usage storage only works on those runtimes.