GitHub - RajatGarga/copilot-openai-server: An HTTP server that exposes OpenAI-compatible API endpoints powered by the GitHub Copilot SDK

Copilot OpenAI-Compatible Server

An HTTP server that exposes OpenAI-compatible API endpoints (/v1/chat/completions, /v1/models), powered by the GitHub Copilot SDK.

This allows you to use GitHub Copilot's agentic capabilities with any client that supports the OpenAI API (e.g., Open WebUI, custom scripts, etc.).

Prerequisites

1. GitHub Copilot CLI must be installed and authenticated.
   • Install via GitHub CLI: gh extension install github/gh-copilot
   • Or ensure the copilot binary is in your PATH.
   • Verify installation: copilot --version
2. Go 1.21+ installed.

Docker

You can run the server using Docker, which includes the GitHub Copilot CLI.

Build the Docker Image

docker build -t copilot-openai-server .

Run the Container

Set your GitHub personal access token with Copilot permissions:

export GH_TOKEN="your_personal_access_token"
docker run -e GH_TOKEN=$GH_TOKEN -p 8080:8080 copilot-openai-server

GitHub personal access token (PAT) must be fine-grained with the Copilot requests permission enabled; see github/copilot-cli#91 for details.

The server will be available at http://localhost:8080.

Versioning note

The Dockerfile pulls the latest Copilot SDK and Copilot CLI at build time instead of pinning specific releases (Dockerfile). If a future SDK or CLI update introduces an incompatible change, the image may break until you pin or update both components together.

Quick Start

1. Build the Server

git clone https://github.com/RajatGarga/copilot-openai-server.git
cd copilot-openai-server
go build -o copilot-server .

2. Run the Server

# Default port is 8080
./copilot-server

# Specify a custom port
./copilot-server -port 9000

API Endpoints

List Models (GET /v1/models)

Lists available models from the Copilot SDK.

curl http://localhost:8080/v1/models

Chat Completions (POST /v1/chat/completions)

Supports standard OpenAI chat completion parameters, including streaming and tool calling.

Basic Request:

curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Hello! How does this server work?"}
    ]
  }'

Streaming Request:

curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [{"role": "user", "content": "Write a short poem about coding."}],
    "stream": true
  }'

Tool Calling:

The server supports defining tools in the OpenAI format. These are passed to the Copilot SDK, which allows the model to "call" them. Note that the client is responsible for executing the tool and sending the result back if needed (standard OpenAI flow).

curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [{"role": "user", "content": "What'\''s the weather in London?"}],
    "tools": [{
      "type": "function",
      "function": {
        "name": "get_weather",
        "description": "Get current weather",
        "parameters": {
          "type": "object",
          "properties": {
            "location": {"type": "string"}
          },
          "required": ["location"]
        }
      }
    }]
  }'

Open WebUI Integration

You can easily use this with Open WebUI:

1. Run this server (./copilot-server).
2. In Open WebUI, go to Settings > Connections > OpenAI.
3. Add a new connection:
   • API Base URL: http://localhost:8080
   • API Key: any-string (not validated by this server, but authentication with Copilot happens via the CLI on the host machine).
4. Save and select a Copilot model to start chatting.

Development

To add this as a dependency in your own Go project:

go get github.com/github/copilot-sdk/go

License

GPL-3.0 License