Local MCP Wrapper for TabFM/TimesFM
Porespellar · reddit · 2026-07-12
The author has wrapped Google's newly released TabFM and TimesFM into an MCP server, allowing local LLMs to directly invoke these tabular and time-series foundation models for classification, regression, and forecasting via Open WebUI, Claude Code, or Codex CLI.
Implementation details:
- Runs as a single Docker container
- Supports dynamic model loading/unloading with a 5-minute TTL
- Currently supports CSV, with plans to support XLS/XLSX/JSON/JSONL later
- Built on PyTorch and strictly CUDA-only
- Targets Nvidia devices with 16GB+ VRAM, such as DGX Spark, 3090, and H100
They also tested it on classic datasets like Iris, California Housing, and Airline Passengers, reporting solid zero-shot performance—such as 94.7% accuracy on Iris and an R² of 0.87 for regression.
Related event: Zer0Fit: Local MCP Server for TabFM and TimesFM(2 posts)→
More from coding & agent
- Microsoft demos intent-classifying agents built with Copilot App and Foundry, deployable across Teams and Outlook — marlene_zw · 2026-10-26
- Looking for a wearable mic with best battery life that feeds recordings to personal agents via MCP — altryne · 2026-10-26
- The identity crisis in software engineering: multiagent systems demand a new problem set — rakyll · 2026-10-26
- Non-CS Beginners Are Shipping Code Through CI/CD Pipelines with AI Help — curious_vii · 2026-10-26
- Ruff creator Charlie Marsh: frontier models find infinite bugs in any software — charliermarsh · 2026-10-26
- Dev testing one skill across Opus 5.5 and Sol 6.1 finds instruction following 'all over the map' — lucasmeijer · 2026-10-26