How to call mcp server in claude code?

Updated October 2026 · How we answer

Short answerYou usually do not call an MCP server by hand in Claude Code. Add it with the claude mcp add command, then ask for a task in plain language, and Claude calls the server's tools when it needs them.

Add the server

Claude Code reads MCP servers from its configuration, and you can register one from the terminal with the claude mcp add command. Give the server a name, then the command or URL it uses.

Choose a scope if the command offers one, such as a project scope or a user scope.

  • Run claude mcp add with a server name
  • Pass the start command or the server URL
  • Add any required environment variables or keys
  • Confirm the server appears in the list

Let Claude use the tools

When a server is connected, its tools become available during your session. You describe the outcome you want, such as looking up records or reading a file, and Claude decides which tool to call and what arguments to send.

Tool calls appear in the session output, and your permission settings control whether you approve each one.

Check that it is connected

If a tool does not show up, check the server list and the status shown in the session. Confirm that the command path is correct and that any API key is available to the process.

Restarting the session often picks up a newly added server.

Common mistakes

  • Trying to type a tool call as a shell command instead of asking for it in a prompt.
  • Adding a server with a wrong command path and never checking the server list.
  • Forgetting that the server may need its own API key in its environment.
From our shopsCaseMorph: Type an idea, see a custom phone case in seconds, then print a one-of-one.