Model Context Protocol (MCP) is an open standard by Anthropic that standardizes how LLMs connect to external data sources and tools. MCP enables: any LLM client to connect to any MCP server using the same protocol. Server types: Resources (static data - files, database records), Tools (functions LLM can call), Prompts (reusable prompt templates). Building an MCP server (Python): from mcp.server import Server; from mcp.server.stdio import stdio_server; app = Server("my-server"); @app.list_tools() async def list_tools(): return [Tool(name="search_database", description="Search company database", inputSchema={"type": "object", "properties": {"query": {"type": "string"}}})]; @app.call_tool() async def call_tool(name, arguments): if name == "search_database": return search_db(arguments["query"]); async with stdio_server() as streams: await app.run(*streams). Connect to Claude Desktop: add server config to claude_desktop_config.json. MCP advantages over custom tool implementations: standardized protocol means one server works with multiple clients (Claude, Zed, Cursor), built-in security model with user approval for tool calls, growing ecosystem of pre-built servers (filesystem, GitHub, PostgreSQL, Slack). Build MCP servers for: internal knowledge bases, CRM data, project management tools, monitoring systems.