copilot-api-proxy

English | 中文

A local proxy that wraps GitHub Copilot as a standard OpenAI / Anthropic-compatible API.
All you need is a GitHub Copilot subscription — no extra API key required.

---

Quick Start

1. Authenticate

./copilot-api-proxy auth

First run only. Authorizes via GitHub Device Flow and caches the token locally.

2. Start the proxy

./copilot-api-proxy start

The proxy is now running at http://127.0.0.1:4142.

3. Configure Claude Code

Open your Claude Code settings.json and add:

{
  "env": {
    "ANTHROPIC_BASE_URL": "http://localhost:4142"
  }
}

Where is settings.json?
macOS: ~/Library/Application Support/Claude/claude_code/settings.json
Linux: ~/.config/Claude/claude_code/settings.json

That's it. Start Claude Code and all Anthropic API requests will be routed through your Copilot subscription.

---

Build from Source

cargo build --release
# Binary: ./target/release/copilot-api-proxy

---

CLI Reference

auth

./copilot-api-proxy auth [flags]

Flag	Description
-f, --force	Force re-authentication, ignoring cached token
--show-token	Print the GitHub Token after auth

start

./copilot-api-proxy start [flags]

Flag	Default	Description
-p, --port	4142	Listening port
--host	127.0.0.1	Listening address. Use 0.0.0.0 to expose to LAN (use with caution)
-v, --verbose	false	Enable DEBUG level logging
-g, --github-token	—	Provide a GitHub Token directly, skip auth flow
-a, --account-type	individual	individual / business / enterprise
--show-token	false	Print GitHub Token and Copilot Token on startup
--proxy	—	HTTP/HTTPS proxy URL, e.g. http://127.0.0.1:7890. Equivalent to setting HTTP_PROXY and HTTPS_PROXY env vars. Required in mainland China to access Claude models.

---

Features

• OpenAI Chat Completions API (/v1/chat/completions)
• Anthropic Messages API (/v1/messages) with streaming
• OpenAI Embeddings API (/v1/embeddings)
• Model listing (/v1/models)
• HTTP/HTTPS proxy passthrough via environment variables

API Endpoints

Endpoint	Description
GET /health	Health check, returns token status
GET /v1/models	List available models
POST /v1/chat/completions	OpenAI Chat Completions (streaming supported)
POST /v1/embeddings	OpenAI Embeddings
POST /v1/messages	Anthropic Messages API (streaming supported)

---

More Examples

Start with an HTTP proxy (required outside mainland China for Claude models)

GitHub Copilot filters available models based on your exit IP. When connecting directly from mainland China, the server omits all claude-* models due to Anthropic's regional restrictions. Routing through an overseas proxy restores the full model list.

./copilot-api-proxy start --port 4142 --proxy http://127.0.0.1:7890

If you start without a proxy and see this warning in the logs:

WARN ⚠️  模型列表中没有 claude-* 模型 ...

it means your current exit IP is in mainland China and Copilot has filtered out Claude models. Add --proxy and restart.

Run as a daemon with PM2

PM2 is a cross-platform process manager that keeps the proxy running in the background, restarts it on crash, and can survive reboots.

Install PM2 (once):

npm install -g pm2

Option 1 — command line (quick start):

pm2 start ./copilot-api-proxy \
  --name copilot-api-proxy \
  --restart-delay 3000 \
  --max-restarts 10 \
  -- start --port 4142 --proxy http://127.0.0.1:7890

Option 2 — config file (recommended for permanent setup):

Create ecosystem.config.js in the same directory as the binary:

module.exports = {
  apps: [
    {
      name: 'copilot-api-proxy',
      script: './copilot-api-proxy',
      args: 'start --port 4142 --proxy http://127.0.0.1:7890',
      restart_delay: 3000,
      max_restarts: 10,
    },
  ],
};

Then start with:

pm2 start ecosystem.config.js

Register for auto-start on login (both options):

pm2 save
pm2 startup   # follow the printed instruction to enable boot persistence

Common commands:

pm2 status                  # check running status
pm2 logs copilot-api-proxy  # tail logs
pm2 restart copilot-api-proxy
pm2 stop copilot-api-proxy
pm2 delete copilot-api-proxy

Note: pm2 status shows N/A for the version column. This is normal — PM2 reads version from package.json, which doesn't exist for a Rust binary. It has no effect on functionality.

./copilot-api-proxy start --github-token ghp_xxxxxxxxxxxx

Use Claude Code with a one-off environment variable

ANTHROPIC_BASE_URL=http://127.0.0.1:4142 claude

Use with an OpenAI-compatible client

from openai import OpenAI

client = OpenAI(
    base_url="http://127.0.0.1:4142/v1",
    api_key="unused",
)

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)

Test with curl

curl http://127.0.0.1:4142/v1/messages \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-3.5-sonnet",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello!"}]
  }'