🤖 Copilot OpenAI API (Go Edition)

A high-performance Go proxy server that turns GitHub Copilot's chat completion/embeddings capabilities into an OpenAI-compatible API service, with Anthropic compatibility and robust security.

I wouldn't call this a "fork" but I would say it's heavily inspired by copilot-openai-api. If you want to use Python instead of Go then you should check out that project.

✨ Key Features

🚀 Advanced Integration

• Seamless GitHub Copilot chat completion API proxy (/v1/chat/completions)
• Embeddings API proxy (/v1/embeddings)
• Anthropic API compatibility (/v1/messages, experimental)
• Real-time streaming response support
• High-performance, concurrent request handling

🔐 Security & Reliability

• Secure authentication middleware (Bearer token)
• Automatic Copilot token management and refresh
• Built-in CORS support for web applications
• Clear error handling (401, 403, etc.)

💻 Universal Compatibility

• Cross-platform config auto-detection (Windows, Unix, macOS)
• Docker containerization ready
• Flexible deployment and configuration

---

🚀 Prerequisites

• Go 1.21+ (https://golang.org/dl/)
• GitHub Copilot subscription
• GitHub authentication token (see below)

---

📦 Installation

1. Clone the repository:

git clone https://github.com/your-org/copilot-openai-api-go.git
cd go-copilot-api

2. Build the binary:

go build -o bin/go-copilot-api ./cmd/go-copilot-api

---

⚙️ Configuration

Set up environment variables (or use a .env file):

Variable	Description	Default
COPILOT_TOKEN	Required. API access token for authentication.	Randomly generated
COPILOT_OAUTH_TOKEN	Copilot OAuth token (auto-detected if not set)	(auto)
COPILOT_SERVER_PORT	Port to listen on (e.g. 8080 for :8080)	9191
CORS_ALLOWED_ORIGINS	Comma-separated list of allowed CORS origins	*
DEBUG	Enable debug logging	false
DEFAULT_MODEL	Default model to use if not specified in request	(none)

Copilot OAuth Token Auto-Detection:

• If COPILOT_OAUTH_TOKEN is not set, the app will look for your Copilot config:
  • Unix/macOS: ~/.config/github-copilot/apps.json
  • Windows: %LOCALAPPDATA%/github-copilot/apps.json
• The first available oauth_token will be used.

How to get a valid Copilot configuration?

• Install any official GitHub Copilot plugin (VS Code, JetBrains, Vim, etc.), sign in, and the config files will be created automatically.

---

🖥️ Local Run

Start the server:

go run ./cmd/go-copilot-api

or

bin/go-copilot-api

---

🔄 Making API Requests

Chat Completions

curl -X POST http://localhost:9191/v1/chat/completions \
  -H "Authorization: Bearer your_access_token_here" \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [{"role": "user", "content": "Hello, Copilot!"}]
  }'

Embeddings

curl -X POST http://localhost:9191/v1/embeddings \
  -H "Authorization: Bearer your_access_token_here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "copilot-text-embedding-ada-002",
    "input": ["The quick brown fox", "Jumped over the lazy dog"]
  }'

Anthropic API Compatibility (Experimental)

curl -X POST http://localhost:9191/v1/messages \
  -H "Authorization: Bearer your_access_token_here" \
  -H "Content-Type: application/json" \
  -d '{ ...Anthropic API message format... }'

List Available Models

Fetch the list of available models and their capabilities:

curl -X GET http://localhost:9191/v1/models

• This endpoint does not require authentication.
• The models list is fetched from GitHub's model catalog API at server startup and periodically refreshed (every 6 hours).
• The response is a JSON array of model objects, including id, name, summary, and more.
• Use the "id" field (e.g., "gpt-5-mini", "gpt-4o-mini-2024-07-18") as the "model" value in your requests.

Example: Find the correct model name for GPT-5 Mini

curl -s http://localhost:9191/v1/models | jq '.[] | select(.id | test("5-mini"))'

---

Default Model

You can set a default model for all requests by adding to your .env or environment:

DEFAULT_MODEL=gpt-5-mini

If a client request does not specify a "model" field, this value will be used automatically for /v1/chat/completions, /v1/embeddings, and /v1/messages.

• If DEFAULT_MODEL is not set, and the client omits "model", no model is sent to Copilot (Copilot will auto-select).
• If the client provides a "model", that value is always used as-is.

Example .env:

COPILOT_TOKEN=your_token_here
DEFAULT_MODEL=gpt-5-mini
COPILOT_SERVER_PORT=9191

🔌 API Reference

POST /v1/chat/completions

• Proxies requests to GitHub Copilot's Completions API.
• Headers: Authorization: Bearer <your_access_token>, Content-Type: application/json
• Body: Must include "messages". You may include "model" (see /v1/models for valid values). If omitted and DEFAULT_MODEL is set, it will be injected.
• Response: Streams responses directly from Copilot (supports streaming and non-streaming).

POST /v1/embeddings

• Proxies requests to Copilot's Embeddings API.
• Headers: Authorization: Bearer <your_access_token>, Content-Type: application/json
• Body: Must include "input". You may include "model" (see /v1/models). If omitted and DEFAULT_MODEL is set, it will be injected.
• Response: JSON from Copilot's embeddings API.

POST /v1/messages

• Converts Anthropic API format to Copilot chat completion format.
• Headers: Authorization: Bearer <your_access_token>, Content-Type: application/json
• Body: Anthropic-compatible. You may include "model" (see /v1/models). If omitted and DEFAULT_MODEL is set, it will be injected.
• Response: Anthropic API-compatible response.

Note: Claude Code/Anthropic compatibility is currently untested. If you use Claude Code or Anthropic clients and encounter issues, we would appreciate any PRs or feedback to help improve support!

GET /v1/models

• Returns a list of available models and their capabilities.
• No authentication required.
• Response: JSON array of models as provided by GitHub's model catalog API.
• Tip: Use the "id" field as the "model" value in your requests.

---

🔒 Authentication

• Set COPILOT_TOKEN in your environment.
• Include in request headers:

  Authorization: Bearer your_access_token_here
  

---

⚠️ Error Handling

• 401: Missing/invalid authorization header
• 403: Invalid access token
• Other errors are propagated from GitHub Copilot API

---

🛡️ Security Best Practices

• Configure CORS for your specific domains (default: *)
• Safeguard your COPILOT_TOKEN and GitHub OAuth token
• Built-in token management with concurrent access protection

---

🧪 Experimental Features

• Anthropic API compatibility (/v1/messages)

---

🗂️ Project Structure

go-copilot-api/
├── cmd/
│   └── go-copilot-api/
│       └── main.go         # Application entrypoint
├── internal/
│   ├── api/                # HTTP handlers and routing
├── pkg/
│   └── config/             # Configuration loading
├── test/                   # Test files
├── LICENSE                 # YO LICENSE
├── go.mod                  # Go module definition
├── go.sum                  # Go module checksums
└── README.md               # This file

---

🧪 Testing

Run all tests:

go test ./...

---

📄 License

See LICENSE — YO LICENSE, Version 1.0, February 2025 Copyright (C) 2025 Travis Peacock