Guides/Advanced Topics
advancedtechnical

MCPs (Model Context Protocol)

Understanding and using MCP servers

What is MCP?

Model Context Protocol (MCP) is a standard for connecting AI models to external tools and data sources. It lets AI assistants interact with databases, APIs, file systems, and more.

Why MCP Matters

Without MCP, AI assistants are limited to what's in their context window. With MCP, they can:

  • Query your database in real-time
  • Access your file system
  • Call external APIs
  • Interact with services like GitHub, Slack, or Notion
  • Execute code and scripts

How MCP Works

MCP uses a client-server architecture:

  • MCP Server — Exposes tools and resources (runs locally or remotely)
  • MCP Client — The AI application that connects to servers (e.g., Claude Desktop)
  • Protocol — JSON-RPC based communication standard

Popular MCP Servers

ServerPurpose
filesystemRead/write local files
githubInteract with GitHub repos, issues, PRs
postgresQuery PostgreSQL databases
slackRead/send Slack messages
brave-searchWeb search capabilities

Setting Up MCP in Claude Desktop

claude_desktop_config.json
{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/path/to/allowed/directory"
      ]
    }
  }
}

MCP is still evolving. Check the official documentation for the latest servers and configuration options.