Summary
You now have a repeatable way to connect an agent powered by the GitHub Copilot harness to tools from a custom or internal service through a Model Context Protocol (MCP) server, and to verify which of those tools the agent calls.
What you learned
In this module, you learned how to:
- Choose an MCP server over a connector or workflow when an MCP server exposes tools for a service that no prebuilt connector covers.
- Add an MCP server to an agent (preview) with a name, description, server URL, and the authentication the server requires, such as an API key, or a client ID and secret.
- Turn on only the MCP tools that the agent's job needs, keep the number of attached servers small, and edit a server's entry to refresh its tool list after the server changes.
- Check the tool, inputs, and result of an MCP tool call in the activity trace, and troubleshoot authentication failures, missing or outdated tools, and tools the agent doesn't call.
Learn more
- Add a Model Context Protocol (MCP) server to your agent as a tool (GitHub Copilot preview)
- Manage and delete tools in an agent (GitHub Copilot)
- Extend your agent with Model Context Protocol
- Use agent tools to extend, automate, and enhance your agents
- Create a new Model Context Protocol (MCP) server
- Use the activity trace to debug your agent
- Configure agent details and instructions