CrewAI Tools provides a framework for extending CrewAI agents with a variety of pre-built and custom tools, enabling them to interact with external systems and perform complex tasks. It targets developers building AI-powered autonomous agents who need to integrate functionalities like file management, web scraping, database access, and AI model interactions.

How It Works

The library offers a collection of ready-to-use tools and a flexible system for creating custom tools, either by subclassing BaseTool or using the @tool decorator. A key feature is its support for the Model Context Protocol (MCP), allowing agents to dynamically discover and utilize thousands of tools exposed by MCP servers. This enables agents to access a vast ecosystem of functionalities without explicit integration for each one.

Quick Start & Requirements

• Primary install: pip install crewai[tools]
• MCP support: pip install crewai-tools[mcp]
• Dependencies: Python 3.12+ recommended for MCP.
• Documentation: https://docs.crewai.com/
• Examples: https://github.com/CrewAIInc/crewAI-examples

Highlighted Details

• Extensive library of tools: File management, web scraping (Selenium), database (PostgreSQL, MySQL), API integrations (Serper, EXA), and AI-powered tools (DALL-E, Vision).
• Model Context Protocol (MCP) integration for dynamic tool discovery from external servers.
• Two methods for custom tool creation: subclassing BaseTool or using the @tool decorator.
• Tools are designed for simplicity, flexibility, and rapid integration into CrewAI agent workflows.

Maintenance & Community

• Active community support via Discourse.
• Contributions welcomed via pull requests.

Licensing & Compatibility

• License: MIT License.
• Compatibility: Designed to work with the CrewAI framework.

Limitations & Caveats

• MCP integration currently only supports tools, not other primitives like prompts or resources.
• MCP tool execution returns only the first text output (.content[0].text).
• Users must trust the source of MCP servers due to potential code execution (STDIO) or injection vulnerabilities (SSE).