github-copilot-api-vscode

GitHub Copilot API Gateway

Use GitHub Copilot, Gemini, Ollama & any VS Code AI — like any other API.
One extension. Zero API keys. Auto-discovers every model. Works with LangChain, Cursor, Aider, and 50+ tools.

---

⚡ Quick Start

# 1. Install from VS Code Marketplace (search "GitHub Copilot API Gateway")
# 2. Click "Start Server" in the sidebar
# 3. Done. Your local API is ready.

Use it like OpenAI:

from openai import OpenAI

client = OpenAI(
    base_url="http://127.0.0.1:3030/v1",
    api_key="anything"  # No real API key needed
)

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

🔒 Your prompts stay private. All AI requests are processed entirely on your machine through VS Code’s Copilot extension — your prompts and API keys never leave your machine. The extension collects anonymous, aggregate usage metrics (server start/stop events, command usage, error types) to help improve reliability. This telemetry is fully optional and respects your VS Code telemetry.telemetryLevel setting — set it to off to disable all data collection. The full source code is available on GitHub for you to audit.

---

✅ What You Can Do

• Use Cursor with Copilot, Gemini, or Ollama as the backend
• Run LangChain agents without paying for OpenAI
• Power AutoGPT and CrewAI agents locally
• Query data with PandasAI and LlamaIndex
• Build chatbots with Botpress, Chainlit, or Rasa
• Pair program with Aider or Open Interpreter
• Generate code with GPT Engineer or Sweep
• Connect Clawdbot to power AI on WhatsApp, Telegram, Discord & Slack
• Mix and match models — route one request to GPT-4o, the next to Gemini, all on localhost:3030

If it speaks OpenAI, it works with this gateway.

🎯 Cursor tip: If Cursor bypasses your custom base URL for model names like gpt-4o, pick a copilot-* model alias from GET /v1/models (for example copilot-gpt-4o). These aliases are provided specifically to prevent Cursor model-name interception.

🔀 Multi-Provider: One Endpoint, Every Model

This gateway auto-discovers all AI models registered in VS Code — not just Copilot.

Provider	How to enable	Example models
GitHub Copilot	Install GitHub Copilot extension	gpt-4o, claude-3.5-sonnet, o3-mini
Google Gemini	Install Gemini Code Assist extension	gemini-2.0-flash
Ollama	Install an Ollama VS Code extension	Local models (llama3, mistral, etc.)
Any other	Any extension using vscode.lm API	Automatically discovered

All models are exposed at GET /v1/models and routable via the model field in any request. No config changes needed — just install the VS Code extension and the models appear.

🤖 Using with Clawdbot (Moltbot)

Clawdbot is an open-source AI assistant that brings LLMs to messaging platforms like WhatsApp, Telegram, Discord, and Slack. Since it supports any OpenAI-compatible API endpoint, you can use this gateway to power Clawdbot with GitHub Copilot:

1. Start the Copilot API Gateway in VS Code
2. Expose the API (set host to 0.0.0.0 if Clawdbot runs on another machine)
3. Configure Clawdbot to use your gateway:

# In your Clawdbot config
llm:
  provider: openai
  base_url: http://YOUR-IP:3030/v1
  api_key: your-optional-api-key  # or "anything" if auth is disabled
  model: gpt-4o

Now your WhatsApp/Telegram/Discord/Slack bots are powered by Copilot! 🚀

🌐 Internet Access via Cloudflare Tunnels

Access your Copilot API from anywhere — your phone, tablet, or share with friends:

1. Enable authentication (set an API key in Security settings)
2. Start the server
3. Click “Go Live” in the Internet Access section
4. Get a public *.trycloudflare.com URL instantly

Free, no Cloudflare account needed! Perfect for:

• Accessing from your phone outside your home network
• Sharing with friends or teammates temporarily
• Remote development without VPN setup

---

🖥️ Run as a Background Service

The gateway runs inside VS Code’s extension host, so VS Code must be running — but it doesn’t need to be visible. Here’s how to run it like a background service:

Prerequisites: Enable autoStart so the server launches automatically:

{ "githubCopilotApi.server.autoStart": true }

macOS:

# Launch VS Code minimized (gateway starts automatically)
open -a "Visual Studio Code" --args --disable-workspace-trust

# Or run completely headless via CLI
nohup code --disable-workspace-trust &>/dev/null &

Windows (PowerShell):

# Launch minimized
Start-Process code -ArgumentList "--disable-workspace-trust" -WindowStyle Minimized

# Or add to Task Scheduler for start-on-boot:
# Program: code
# Arguments: --disable-workspace-trust

Linux:

# Run in background
nohup code --disable-workspace-trust &>/dev/null &

# Or create a systemd user service (~/.config/systemd/user/copilot-api.service):
# [Service]
# ExecStart=/usr/bin/code --disable-workspace-trust
# Restart=on-failure
# Then: systemctl --user enable --now copilot-api

Note: VS Code must remain running in the background. If you close VS Code, the gateway stops. The autoStart setting ensures the API server starts the moment VS Code opens — no manual clicks needed.

🔌 API Endpoints

Endpoint	Method	Description
/v1/chat/completions	POST	Chat completions (streaming supported)
/v1/completions	POST	Legacy completions API (streaming supported)
/v1/responses	POST	OpenAI Responses API (2026 spec, streaming supported)
/v1/models	GET	List all available models
/v1/tools	GET	List available tools (VS Code + MCP)
/v1/tools/call	POST	Execute a tool directly
/v1/mcp/servers	GET	List connected MCP servers
/v1/messages	POST	Anthropic Claude-compatible endpoint
/v1beta/models/:model:generateContent	POST	Google Gemini-compatible endpoint
/health	GET	Server health check
/docs	GET	Interactive Swagger UI

---

⚙️ Configuration

Customize in VS Code Settings (githubCopilotApi.*):

{
  "githubCopilotApi.server.port": 3030,
  "githubCopilotApi.server.host": "127.0.0.1",
  "githubCopilotApi.server.apiKey": "",
  "githubCopilotApi.server.autoStart": false
}

Setting	Default	Description
server.port	3030	Local server port
server.host	127.0.0.1	Bind address (use 0.0.0.0 for LAN access)
server.apiKey	(empty)	Optional Bearer token for authentication
server.allowedHosts	[]	Additional exact Host names accepted by HTTP and WebSocket APIs through a reverse proxy
server.allowedOrigins	[]	Exact browser origins allowed to call the APIs cross-origin
server.autoStart	false	Start server when VS Code opens

For advanced options (TLS, IP allowlisting, audit logging), see CONTRIBUTING.md.

HTTP and WebSocket requests validate Host and Origin. Browser requests are same-origin by default; add an exact origin such as https://app.example.com to server.allowedOrigins only when a browser app needs cross-origin access. For a reverse proxy, add its exact gateway hostname to server.allowedHosts. JSON POST requests must use Content-Type: application/json. The former server.webSocketAllowedHosts and server.webSocketAllowedOrigins settings remain supported as deprecated aliases. Native WebSocket clients can send Authorization: Bearer <key> or x-api-key; browser clients can pass api-key.<key> as a WebSocket subprotocol. The credential subprotocol is not returned in the handshake response.

---

📦 Requirements

• VS Code 1.95+
• GitHub Copilot Chat extension (must be signed in)

---

📖 Documentation

• Official Docs
• Contributing Guide
• Changelog

---

🛡️ Trust & Transparency

This project is 100% free and open-source under the MIT License.

• 🔍 Full source code available on GitHub
• 📊 Anonymous telemetry — collects aggregate extension health metrics only (no prompts, no file paths, no PII). Respects your VS Code telemetry.telemetryLevel setting — disable via "telemetry.telemetryLevel": "off" in VS Code settings.
• 🔒 Prompts stay local — your AI requests and API keys never leave your machine
• ✅ Community-driven — contributions welcome via CONTRIBUTING.md
• 🛡️ Security — report vulnerabilities via SECURITY.md

Disclaimer: This extension is an independent, community-built project and is not affiliated with GitHub, Microsoft, or OpenAI. It leverages your existing GitHub Copilot subscription. Use responsibly.

Built with ❤️ by Suhaib Bin Younis