Zer0Fit wraps Google's TabFM and TimesFM into a local MCP for zero-shot ML

Porespellar · reddit · 2026-09-27

An ML grad student open-sourced Zer0Fit, which wraps two Google Research zero-shot foundation models — TabFM (tabular classification/regression) and TimesFM (time-series forecasting) — behind a FastAPI service packaged as a dockerized MCP that plugs into any MCP-friendly LLM like Open WebUI.

Usage: connect the MCP to your LLM, hand it CSV/tabular/time-series data with a natural-language request, and the LLM picks the right Google model, runs the task in context, and returns results — no model training needed.

The author is candid that this loses to a data scientist's custom model, but for tasks that don't demand high precision or speed it gets you a ballpark answer. All heavy lifting is done by Google's models; he just wrapped them for local serving via MCP, and warns against using it where output matters.

Original post →

More from coding & agent

coding & agent channel →