Tools

Executable functions that MCP servers expose to clients, enabling LLMs to perform actions, interact with external systems, and modify state in the real world.

Tools

Tools are executable functions that MCP servers expose to clients, enabling LLMs to perform actions and interact with external systems. Tools are model-controlled, designed for AI models to automatically invoke with human approval.

Tool Structure

Component	Purpose
Name	Unique identifier for the tool
Description	Human-readable explanation of functionality
Input Schema	JSON Schema defining required parameters
Annotations	Metadata about tool behavior and side effects

Tool Categories

System Operations - Execute shell commands, manage files
API Integrations - Create GitHub issues, send emails
Data Processing - Analyze CSV files, transform data
Database Operations - Query records, update tables

Tool Annotations

Tools include helpful metadata hints:

{
  "annotations": {
    "title": "Human-friendly display name",
    "readOnlyHint": true,      // Doesn't modify environment
    "destructiveHint": false,   // May perform destructive updates  
    "idempotentHint": true,     // Repeated calls have no additional effect
    "openWorldHint": false      // Interacts with external entities
  }
}

Discovery and Execution

// List available tools
{
  "method": "tools/list"
}

// Execute a tool
{
  "method": "tools/call", 
  "params": {
    "name": "create_file",
    "arguments": {
      "path": "/tmp/example.txt",
      "content": "Hello world"
    }
  }
}