2 posts tagged with "mcp"

Today we're excited to announce significant enhancements to Agent Runtimes — our open-source framework for building AI agents that can connect to any tool, any model, and any interface.

The first user is the Jupyter AI Agents extension, which brings intelligent agents directly into JupyterLab notebooks. But Agent Runtimes is designed as a general-purpose framework that works in any environment.

These updates introduce industry-standard transport protocols, rich UI extensions, and first-class Model Context Protocol (MCP) support — all designed to make it easier than ever to build production-ready AI agents that do real work.

I recently had the privilege of presenting at JupyterCon 2025, where I shared our work on the Jupyter MCP Server that aims to standardize how AI agents interact with Jupyter notebooks through the Model Context Protocol (MCP).