# fast-agent MCP server

> Lets an AI client build, chat with, and evaluate LLM agents and workflows that connect to MCP servers.

- Repository: https://github.com/evalstate/fast-agent
- Page: https://hostedmcp.org/servers/evalstate/fast-agent
- Transport: —
- Stars: 3965
- License: Apache-2.0

## FAQ

### How do I start an interactive fast-agent session?

Install uv and run `uvx fast-agent-mcp@latest -x`, or install the tool and run `fast-agent go`.

### How do I connect fast-agent to an MCP server?

Use the `/connect` command, which supports stdio or streamable HTTP servers with OAuth.

### Does fast-agent support OAuth for remote servers?

Yes, OAuth is supported with KeyRing storage for secrets, managed via the `fast-agent auth` command.

### Can fast-agent connect to local models like llama.cpp?

Yes, use the generic provider or run `fast-agent model llamacpp` to auto-generate configuration.

### Can fast-agent run as an MCP server itself?

Yes, running with `--transport http --port 8001` exposes the agent as an MCP server.