Getting Started with MCP Servers: Connect Your AI Tools to the World

What is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external tools and data sources. Instead of just generating text, an AI with MCP access can query databases, run browser tests, call APIs, and interact with real systems.

Think of MCP servers as plugins for your AI tools. Each server exposes specific capabilities — and your AI assistant can use them when it needs to.

What can you do with MCP servers?

Without MCP, your AI assistant is limited to what it can do with text. With MCP, it can:

• Query your database directly instead of you copy-pasting schema information
• Run browser tests to verify UI changes actually render correctly
• Search documentation for 9,000+ libraries to give accurate API advice
• Create files in formats like Excel, PDF, and PowerPoint
• Access external APIs for research, analytics, and automation

How do you set up an MCP server?

The setup is the same across tools — you add a JSON configuration that tells your AI tool how to start the MCP server.

1. Playwright MCP (browser automation)

This lets your AI control a browser for testing and verification:

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": ["-y", "@playwright/mcp"]
    }
  }
}

2. Docfork (library documentation)

Gives your AI access to up-to-date docs for thousands of libraries:

{
  "mcpServers": {
    "docfork": {
      "command": "npx",
      "args": ["-y", "docfork"]
    }
  }
}

3. Agent Shelf MCP (agent registry)

Search and download agents directly from your AI tool:

{
  "mcpServers": {
    "agentshelf": {
      "type": "http",
      "url": "https://www.agentshelf.dev/api/mcp"
    }
  }
}

Where does the MCP config file go?

The config file location depends on your tool:

• Claude Code — .mcp.json in your project root or ~/.mcp.json for global
• Claude Desktop — claude_desktop_config.json in ~/Library/Application Support/Claude/
• Cursor — .cursor/mcp.json in your project root
• Windsurf — ~/.codeium/windsurf/mcp_config.json

How do MCP servers work with AI agents?

MCP servers become even more powerful when combined with agents. An agent definition can reference specific MCP servers, telling users exactly what to install and why.

For example, our Data Analyst agent references ClickHouse, Elasticsearch, Redis, and Vizro MCP servers — each with copy-paste configuration blocks. Install the agent, install the MCP servers it needs, and you have a fully-equipped data analysis specialist.

Browse agents on Agent Shelf to find ones with MCP integrations already configured.

Learn more

• Agent Shelf MCP Server page — Full setup guide for the Agent Shelf MCP server
• Model Context Protocol spec — The official MCP documentation
• Glama MCP Directory — Browse 21,000+ MCP servers