The Model Context Protocol (MCP) has everyone including myself fired up because of the way it standardizes connecting tools to LLMs. With it, creating powerful AI agents that can do practically anything for you has become MUCH more accessible.
However, as powerful as it is, it sucks to be limited to using MCP servers (these sets of AI agent tools) within only the existing clients that support MCP like Claude Desktop, Windsurf, and Cursor. What if you want to use MCP servers as tools for AI agents you build yourself?
That’s what I’m going to show you how to do in this video - integrate MCP servers with custom AI agents. We’ll use Pydantic AI as our AI agent framework, but this process can be done with really any framework or with no framework at all!
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Try CosmicUp today to get all of the best LLMs in a single place for a very low price:
https://cosmicup.me/
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The code in this video for the Pydantic AI and MCP integration can be found here:
https://github.com/coleam00/ottomator-agents/tree/main/pydantic-ai-mcp-agent
Previous video covering everything you need to know about MCP at a high level:
https://youtu.be/v_6EXt6T83I
MCP documentation:
https://modelcontextprotocol.io/introduction
List of MCP servers on GitHub:
https://github.com/modelcontextprotocol/servers
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
00:00 - Don't Sleep on MCP and Custom AI Agents
01:25 - Why MCP + Custom AI Agents?
03:03 - Resources I Used to Connect MCP to My Agents
03:44 - Step by Step - Integrate MCP with Your AI Agents
06:29 - The Power of MCP + Pydantic AI
09:03 - CosmicUp
10:59 - Covering the Basic Example in the Repo
11:33 - Building Our Custom MCP Client
17:47 - Final Demo
19:22 - How This Integration can be Extended
20:54 - Outro
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Join me as I push the limits of what is possible with AI. I'll be uploading videos at least two times a week - Sundays and Wednesdays at 7:00 PM CDT!