Set up MCP servers (beginner guide)

MCP (Model Context Protocol) lets your agent use external tools in a consistent, auditable way. Start with one trusted server and verify tool calls before adding more.

1

Understand the parts

  • Client: your agent runtime.
  • Server: tool provider that exposes capabilities.
  • Transport: stdio or HTTP endpoint between client and server.
2

Configure your first server

  1. Add server command/URL to your agent MCP settings.
  2. Add required environment variables and API keys as secrets.
  3. Restart your agent so it can discover tools.
3

Validate tool calls safely

  • Ask your agent to list available MCP tools.
  • Run read-only commands first.
  • Enable write actions only after logs and permissions look correct.

Sign up

Welcome to ClawNode

Pick a plan: OpenClaw (bring your own API key) or a Node (Ollama + open-source LLM preinstalled).

or

By selecting Agree and continue, I agree to ClawNode's Terms of Service and Privacy Policy.

or

Already have an account? Log in