The Joe Reis Show The Surprising Rise of FastMCP, a 1M+ Download/Day Hit w/ Jeremiah Lowin
Oct 29, 2025
Jeremiah Lowin, founder of Prefect and creator of FastMCP, dives into the seismic shifts in the AI and data landscape. He shares the fascinating origin of FastMCP as a Pythonic layer for Anthropic's Model Context Protocol, which skyrocketed to over a million downloads daily after OpenAI and Google support. Lowin emphasizes MCP's role as a 'USB-C for AI agents' and highlights its unexpected adoption by internal data teams seeking governable interfaces. Tune in for insights on the intersection of AI, data workflows, and the challenges of open-source maintenance.
AI Snips
Chapters
Transcript
Episode notes
Orchestration Roots Inform AI Tools
- Prefect began as an orchestration platform for data science to make workflows feel natural for PyData users.
- The same engineering DNA later enabled Prefect to build MCP tooling that serves overlapping data and AI audiences.
Weekend Project Turned Official SDK
- Jeremiah built FastMCP as a more Pythonic, high-level interface to Anthropic's MCP after finding the SDK rough to use.
- Anthropic later incorporated FastMCP into their official SDK after Jeremiah relicensed and collaborated with them.
Evolving From SDK To App Primitives
- FastMCP shifted from focusing on server DX to higher-level primitives for building MCP applications.
- The project aimed to answer what an MCP application should look like for autonomous agents rather than human users.