@hyorman/copilot-proxy-cli

OpenAI-compatible proxy server for GitHub Copilot. Use Copilot models through standard OpenAI API endpoints from any HTTP client.

This project is not affiliated with GitHub, Microsoft, or OpenAI. It is intended for research, experimentation, and local prototyping rather than production use.

Quick Start

npx @hyorman/copilot-proxy-cli

If GITHUB_TOKEN is unset, the CLI falls back to the Copilot SDK's device code flow on first run.

Test it:

curl http://localhost:3000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "gpt-4o", "messages": [{"role": "user", "content": "Hello!"}]}'

Installation

Run without installing (recommended):

npx @hyorman/copilot-proxy-cli

Install globally:

npm install -g @hyorman/copilot-proxy-cli
copilot-proxy

From source:

git clone https://github.com/hyorman/copilot-proxy.git
cd copilot-proxy
npm install && npm run build
node packages/cli/out/main.js

Configuration

Environment variables

| Variable | Description | Default | | --- | --- | --- | | GITHUB_TOKEN | GitHub personal access token with Copilot access (optional) | — | | PORT | Server listening port | 3000 | | API_TOKENS | Comma-separated bearer tokens for request authentication | — | | DATA_DIR | Directory for persisted state (tokens, assistants, skills) | ~/.copilot-sdk-proxy | | CLI_PATH | Path to Copilot CLI executable | — | | CLI_URL | URL of an existing CLI server to connect to | — |

.env file support

The CLI automatically loads a .env file from the working directory via Node's process.loadEnvFile() on supported runtimes. Create a .env file to avoid exporting variables each time:

GITHUB_TOKEN=ghp_...
PORT=3001
API_TOKENS=my-secret-token-1,my-secret-token-2

If GITHUB_TOKEN is omitted, the CLI will prompt for device code authentication when the SDK needs to sign in.

Token management

The CLI includes subcommands for managing persistent API tokens. Tokens are stored in DATA_DIR (default ~/.copilot-sdk-proxy/api-tokens.json).

Create a token:

copilot-proxy token create my-token

List tokens:

copilot-proxy token list

Remove a token:

copilot-proxy token remove my-token

When both API_TOKENS env var and persisted tokens exist, they are merged. If no tokens are configured at all, authentication is disabled and the server accepts all requests.

API Endpoints

| Method | Path | Description | | --- | --- | --- | | GET | /v1/models | List available models | | POST | /v1/chat/completions | Chat completions (streaming + non-streaming) | | POST | /v1/responses | Model responses | | — | /v1/assistants/... | Assistants API | | — | /v1/skills/... | Skills API |

Usage Examples

Chat completion

curl -X POST http://localhost:3000/v1/chat/completions \
  -H "Authorization: Bearer cpx_your_token" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [{"role": "user", "content": "Hello!"}],
    "stream": false
  }'

Streaming

curl -X POST http://localhost:3000/v1/chat/completions \
  -H "Authorization: Bearer cpx_your_token" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [{"role": "user", "content": "Hello!"}],
    "stream": true
  }'

Requirements

• Node.js 22+
• A GitHub account with Copilot access

License

MIT

See the main repository for full documentation including the Skills API and Assistants API.