Labsco
MCP SERVER

FAIM Time-Series forecasting

by S-FM

Run point or probabilistic time-series forecasts on FAIM's hosted models, choosing between Chronos2 and TiRex per call.

Physical Sciences & Scientific ComputingVerified
Summary
Forecasting becomes a call the assistant makes mid-conversation rather than a notebook you open.

The series travels in the call as the x parameter and the model is chosen by name, so switching from TiRex to Chronos2 is a parameter change rather than a new environment. TiRex is univariate only, which puts anything multivariate on Chronos2. The two tools cover choosing a model and running it; storing or fetching the series is not part of the surface.

What it is

A two-tool front end for the FAIM forecasting platform: one call lists the models on offer, the other runs the forecast.

What you get
  • Point forecasts for a single value and probabilistic forecasts with confidence intervals, over a horizon you set.
  • Quantiles on request, so the output can be a distribution rather than one number.
  • Univariate and multivariate series both accepted, with Chronos2 the default for multivariate and TiRex restricted to univariate.
  • A model listing that reports supported output types and features before you commit a run.
Requirements

A FAIM platform account and its key, supplied as FAIM_API_KEY.

Setup effort

One command plus a key — npm install -g @faim-group/mcp, then supply credentials