What is mcp-cli?

mcp-cli is a command-line inspector for the Model Context Protocol (MCP). It is run via npx and requires Node.js/npm.

How to use mcp-cli?

You can run npx @wong2/mcp-cli without arguments to use the Claude Desktop config file, or specify a config file with -c config.json. Servers can also be started from NPM packages, locally developed scripts, or connected to running Streamable HTTP or SSE servers.

Key features of mcp-cli

- Run MCP servers from various sources
- List Tools, Resources, and Prompts
- Call Tools, Read Resources, and Read Prompts
- OAuth support for SSE and Streamable HTTP servers
- Connect to running servers over SSE or Streamable HTTP

Use cases of mcp-cli

- Inspect and interact with MCP servers directly from the terminal
- Test locally developed MCP servers before integrating them into an application
- Debug existing MCP servers by listing available tools, resources, and prompts

FAQ from mcp-cli

What kinds of MCP servers can be used with mcp-cli?

mcp-cli supports servers defined in a Claude Desktop config file, NPM packages, local scripts (node path/to/server.js), and running servers over Streamable HTTP (--url) or SSE (--sse).

Does mcp-cli support OAuth?

Yes, mcp-cli has OAuth support for SSE and Streamable HTTP servers.

How do I clear stored data like OAuth tokens?

Run npx @wong2/mcp-cli purge to purge stored data.

Can I use mcp-cli without installing anything?

Yes, you can run it directly with npx @wong2/mcp-cli – no global install is required.

What is the connection format for a running server?

Use --url <endpoint> for Streamable HTTP servers or --sse <endpoint> for SSE servers, for example --url http://localhost:8000/mcp.