faim-mcp

Faim MCP Server

FAIM's Chronos2 and TiRex models bring zero-shot time-series forecasting to MCP-compatible AI assistants through faim-mcp. Its tools expose model capabilities and generate point predictions, quantiles, or distribution samples.

Server facts

faim-mcp

Project

Author
S-FM
License
Not specified

Tech

Runtime / language
Not specified
Transports
stdio
Remote available
No
Authentication
Not specified

Activity

GitHub stars
8
Last commit
Data as of

Tags

  • forecasting
  • time series

Overview

Time-series analysis through faim-mcp starts with discovering model capabilities and choosing a forecasting model. The integration connects MCP clients to the FAIM forecasting SDK, exposing list_models for model information and forecast for prediction requests. Chronos2 is the default model and supports multivariate inputs, while TiRex handles univariate series. This gives users of compatible assistants access to zero-shot forecasting through the FAIM platform, with model selection and forecast generation available as separate tools.

Forecast requests specify the model, input observations, prediction horizon, and desired output type. Inputs can be a 1D array representing a single series or a 3D array organized by batch, sequence, and feature. Results can contain point predictions, quantiles with user-selected levels, or samples from the forecast distribution, supporting workflows that need uncertainty estimates alongside predicted values. The server communicates over stdio using JSON-RPC and requires a FAIM API key supplied through the environment.

faim-mcp setup notes

  • faim-mcp records stdio only; confirm a network transport upstream before planning remote access.
  • faim-mcp has no recorded license; confirm reuse terms in its repository.

Related MCP servers