🤖 HarDojo
Log in Sign up

What is MCP?

The open standard that gives AI tools superpowers.

🧪 Try it in the Playground 🎯 Practice this path

MCP — the Model Context Protocol — is an open standard (introduced by Anthropic in late 2024) that lets AI apps connect to external tools and data through a single, consistent interface.

The problem it solves

Before MCP, every AI app integrated with every tool separately: a Claude integration for GitHub, a different one for Slack, another for your database. That's N × M integrations.

MCP makes it one protocol:

AI app  ──MCP──>  filesystem server
   │────MCP──>  GitHub server
   └────MCP──>  your database server

Any MCP client (Claude, Cursor, Windsurf…) can use any MCP server. Add a server once, use it everywhere.

What MCP provides

Three primitives:

  1. Tools — actions the model can call (read_file, create_issue, query_db).
  2. Resources — data the model can read (a file, a schema, a document).
  3. Prompts — reusable prompt templates (a code-review template).

Plus a transport that carries the messages (stdio or HTTP/SSE).

A concrete example

Add a filesystem server to Claude Desktop. Suddenly Claude can:

> Read todo.md and move the completed items to done.md

The model sees the available tools, calls read_file and write_file, and the server executes them — the model never needs its own file-access code.

Why you should care

  • Reuse — one server serves every client.
  • Security — servers run locally, so your tools don't live on the model provider's cloud.
  • Ecosystem — thousands of prebuilt servers exist (GitHub, Slack, Postgres, Playwright, and more).
MCP is to AI tools what USB-C is to hardware: one connector for everything.