MCP Tools

Standardized executable functions that bridge LLMs to local data and third-party APIs.

MCP Tools serve as the functional backbone of the Model Context Protocol (MCP), enabling AI models to move beyond text generation and into active task execution. By implementing a universal JSON-RPC interface, these tools allow developers to expose specific Python or TypeScript functions—such as fetching GitHub pull requests, querying PostgreSQL databases, or searching Slack archives—directly to an LLM. This architecture eliminates the need for custom glue code between different AI clients like Claude Desktop or IDE extensions. Operators can deploy pre-built servers from the MCP ecosystem or build custom tools to give models secure, real-time access to internal infrastructure.

https://modelcontextprotocol.io