Documentation
GitHub Copilot
Configure Milkey for GitHub Copilot in IDE-based Agent mode or Copilot CLI.
GitHub Copilot now supports MCP across IDE and CLI surfaces, but the config shape differs depending on where you use Copilot.
Use the IDE format when you are configuring Copilot Chat in Agent mode. Use the CLI format when you want Milkey available from GitHub Copilot CLI sessions.
Configure GitHub Copilot
The client-specific work here is mostly about where the config lives and which field names the client expects.
- For Copilot Chat in VS Code, Visual Studio, JetBrains, or Xcode, edit the relevant `mcp.json` file from the Copilot MCP UI.
- For Copilot CLI, edit `~/.copilot/mcp-config.json` or use the built-in `/mcp add` flow in interactive mode.
{ "servers": { "milkey": { "url": "https://mcp.vexelityai.com/mcp", "requestInit": { "headers": { "Authorization": "Bearer mk_sk_your_api_key_id_your_secret" } } } }}Copilot IDEs
mcp.json
Copilot CLI
~/.copilot/mcp-config.json
Verify the connection
- In Copilot Chat, switch to Agent mode and confirm Milkey appears in the tools list after the server starts.
- In Copilot CLI, run `/mcp show` to confirm the server exists and is enabled.
- Ask for a specific Milkey-driven task so you can verify tools are actually being used.
Test prompt
Use Milkey to find the best skill for API pagination design in this repository and summarize the recommendations before changing code.Operational tips
- If your organization manages Copilot MCP policies, make sure external MCP servers are allowed before debugging the Milkey config.
- Use one key per environment so IDE and CLI access can be rotated independently.