How to install an MCP server safely
Learn how to choose, configure, and test an MCP server in your AI client.
1. Choose a compatible server
Find a server in the FreeMCP directory and open its original repository. Confirm it supports your AI client and operating system.
2. Read the maintainer's instructions
Check prerequisites, installation commands, supported transport (such as stdio or Streamable HTTP), environment variables, and required permissions. Do not assume example commands work for every project.
3. Inspect security requirements
Review source code and dependencies where possible. Prefer least-privilege credentials, scoped tokens, and a test environment. Never paste secrets into public issues or chat logs.
4. Add the server to your MCP client
Use your client's documented configuration interface. A local server generally requires an executable command and arguments; a remote server may require an endpoint and authentication. Configuration formats differ by client.
5. Restart and verify
Restart or reconnect your AI client if required, confirm the server appears, and test a low-risk action. Check logs and the maintainer's troubleshooting guidance if it fails.
6. Maintain your setup
Keep dependencies updated, review newly requested permissions, and revoke credentials when a server is no longer needed.
For exact configuration examples, consult your client's official MCP documentation and the server's README.