Package Exports

@ascdong/copilot-proxy
@ascdong/copilot-proxy/dist/index.js

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (@ascdong/copilot-proxy) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

copilot-proxy

Expose your GitHub Copilot subscription as OpenAI-, Anthropic-, and Gemini-compatible HTTP endpoints, so any tool that speaks those APIs — Claude Code, Codex CLI, Gemini CLI, custom scripts — can route through Copilot.

Runs on Node.js ≥ 20 (or Bun in development).

Features

OpenAI-compatible: POST /v1/chat/completions, /chat/completions, /v1/responses (streaming + non-streaming), plus GET /v1/models
Anthropic-compatible: POST /v1/messages, /v1/messages/count_tokens (streaming + non-streaming)
Gemini-compatible: POST /v1beta/models/{model}:generateContent / :streamGenerateContent / :countTokens, plus GET /v1beta/models for CLI preflight
Format translation: Anthropic ↔ OpenAI, Gemini ↔ OpenAI, and Responses ↔ Chat Completions for models that don't support /v1/responses natively (e.g. Claude via Copilot)
Web search fallback — when a model doesn't natively support Anthropic's web_search tool, the proxy transparently runs the query via Tavily and injects synthetic server_tool_use / web_search_tool_result blocks
Vision passthrough — image inputs are forwarded to vision-capable Copilot models
GitHub Device Flow OAuth — one-time login, tokens persisted locally
Auto-refreshing Copilot token — the short-lived upstream token is refreshed in the background
Model aliasing — claude-opus-4-6 ↔ claude-opus-4.6, etc.
Token usage tracking — every request logged with input_tokens / output_tokens / cache_read_input_tokens / duration / status
One-shot client configuration — wire Claude Code, Codex, and Gemini CLIs to this proxy without hand-editing settings files
systemd user service — install/uninstall as a background service on Linux/WSL

Install

Requires Node.js ≥ 20.

npm install -g @ascdong/copilot-proxy

Quickstart

# 1. Sign in to GitHub (Device Flow — opens your browser)
copilot-proxy login

# 2. Start the proxy (defaults to http://127.0.0.1:8989)
copilot-proxy start

# 3. (Optional) wire up a client CLI
copilot-proxy config claude   # or: codex | gemini | all

Once running, point any OpenAI/Anthropic/Gemini client at the proxy. Any API key will do — the proxy uses your GitHub Copilot session, not the client-provided key.

CLI

copilot-proxy <command> [options]

Commands:
  start [-p <port>] [-H <host>]        start the proxy server
  login                                sign in to GitHub via Device Flow OAuth
  config <claude|codex|gemini|all>     configure a client tool
         [-o <path>]                   override the settings output path
  usage [-m <YYYY-MM>]                 show token usage statistics
  logs  [-l <n>] [-e] [--model <m>]    show recent request logs
        [-d <YYYY-MM-DD>]
  service <install|uninstall>          manage the systemd user service (Linux/WSL)

Run copilot-proxy --help for the full listing with examples.

`config` — wire up a client

copilot-proxy config claude    # writes ~/.claude/settings.json (+ Windows path under WSL)
copilot-proxy config codex     # writes ~/.codex/config.toml  (Copilot Proxy or AOAI mode)
copilot-proxy config gemini    # writes ~/.gemini/.env + ~/.gemini/settings.json
copilot-proxy config all       # all of the above

Existing settings are merged, not overwritten.

Optional: Web search via Tavily

Edit ~/.copilot-proxy/config.yaml (auto-generated on first run) and fill in the web_search block:

web_search:
  enabled: true
  provider: "tavily"
  tavily_api_key: "tvly-..."   # get one at https://tavily.com

Restart the proxy. When a model rejects Anthropic's web_search tool, the proxy will run the query via Tavily and inject synthetic server_tool_use / web_search_tool_result blocks so the client still gets a normal Anthropic-shaped response. With enabled: false (the default) or a blank key, web_search requests are passed through unchanged and may be rejected by the upstream model.

`logs` — inspect requests

[13:34:22] 200  claude-opus-4.7          /v1/messages              1 in    938 out  21.2s
 │          │    │                        │                         │         │        └─ duration
 │          │    │                        │                         │         └─ output tokens
 │          │    │                        │                         └─ input tokens
 │          │    │                        └─ endpoint
 │          │    └─ model (translated name when aliased)
 │          └─ HTTP status
 └─ request time

Request logs live at ~/.copilot-proxy/logs/requests/YYYY-MM-DD.jsonl.

`service install` (Linux/WSL)

Registers a user-level systemd unit (~/.config/systemd/user/copilot-proxy.service) that runs the proxy on boot/login and restarts on failure. Uses the globally-installed copilot-proxy binary, so it survives package updates.

copilot-proxy service install
systemctl --user status copilot-proxy
journalctl --user -u copilot-proxy -f
copilot-proxy service uninstall

On WSL, systemd must be enabled (systemd=true in /etc/wsl.conf). To keep the service running when no user is logged in: loginctl enable-linger $USER.

Configuration file

Optional YAML at ~/.copilot-proxy/config.yaml. Defaults are fine for most users; override port / address / log level as needed. Set web_search.enabled: true + web_search.tavily_api_key to enable the Tavily fallback. Environment variable GITHUB_TOKEN can be used in place of the interactive login.

References

License

MIT

@ascdong/copilot-proxy