What is mcp?

This is a command-line tool that lets you call MCP servers directly from a terminal. MCP stands for Model Context Protocol, a standard for exposing tools that AI assistants can use, such as searching error logs, querying dashboards, or reading files. This CLI lets you call those same tools without going through an AI assistant, treating them like regular terminal commands. You configure servers once in a config file and then call their tools with a simple command: the server name, the tool name, and any arguments as a JSON string. Output is always JSON, so it can be piped into other tools or used in scripts. The README shows examples like searching for open errors in Sentry, listing Slack channels, or querying a Grafana dashboard. Servers can be local processes that the CLI launches on demand, remote servers accessed over HTTP, or wrappers around any existing command-line tool. For HTTP servers that require authentication, the CLI handles the login flow: it supports OAuth 2.0 (opening your browser to complete the login), manual API tokens, or headers set directly in the config file. Tokens are saved locally so you only need to authenticate once. There is a built-in registry of common MCP servers so you can add well-known ones with a single command rather than writing the config manually. Installation is available via Homebrew, Docker, or pre-built binaries from the releases page. The project is written in Rust and distributed as a single binary with no runtime dependencies.