What Is an MCP Server?
An MCP server is a program that exposes tools and data to AI agents over the Model Context Protocol (MCP), an open standard, so any MCP-compatible client — Claude, Cursor, or a custom agent — can call those tools instead of guessing at the answer.
How it works
The client connects to the server over a transport — streamable HTTP for a hosted server, or stdio for a process the client launches itself — and asks which tools the server offers. Each tool has a name, a description and a typed input schema, so the model can decide when to call it and with what arguments.
The server does the work and returns structured results. The model never needs the whole system in its context window: it asks a precise question and gets a precise answer back.
MCP servers for codebases
A codebase MCP server answers questions an agent cannot answer from the files it happens to have open: what depends on this file, what breaks if it changes, which code is unused, who owns a module. Each of those needs the whole repository parsed into a graph, which is more than fits in any context window.
Mapnostics runs a hosted MCP server over an indexed GitHub repository, with tools such as get_impact, get_dependencies and get_dead_code. Clients connect over streamable HTTP with a Bearer API key; a stdio bridge (npx mapnostics-mcp) covers clients that only speak stdio.
Common questions
What is an MCP server?
A program that exposes tools and data to AI agents over the Model Context Protocol, an open standard, so any compatible client can call those tools instead of guessing.
What is the difference between a local and a hosted MCP server?
A local server runs as a process on your machine and talks to the client over stdio. A hosted server runs remotely and is reached over HTTP, usually with an API key. The protocol is the same; only the transport differs.
How do I connect Claude Desktop or Cursor to a codebase MCP server?
Add the server to the client's MCP configuration — claude_desktop_config.json for Claude Desktop, .cursor/mcp.json for Cursor — with its URL or launch command and any credentials it needs. The client then lists the server's tools and calls them when a question needs one.
Related
- MCP server — Give Claude, Cursor and any MCP-compatible agent grounded answers about your codebase instead of guesses. A full toolset over your real dependency graph — connect in minutes, no vector database to run yourself.
- Codebase context for AI agents — glossary
- Dependency graph — glossary
- Codebase intelligence — glossary
See it in your own codebase.
Analyze my codebase →No credit card required · Free tier available