forecast-mcp 0.1.1

An MCP server that gives any AI agent time-series forecasting superpowers (TimesFM 2.5 + statistical baselines).

forecast-mcp

Give any AI agent time-series forecasting superpowers.

An MCP server that lets Claude Code, Claude Desktop, Cursor, or any MCP client forecast a series of numbers — sales, traffic, usage, costs — and reason about the result. Powered by Google's TimesFM 2.5 foundation model, with a zero-dependency statistical baseline so it works the moment you install it.

Illustrative — statistical baseline backend. Your agent calls forecast(values=[...], horizon=6, quantiles=[0.9]) and gets back these numbers plus a plain-language summary.

Why

LLM agents can read, write, and run code — but they can't see the future. This gives them a clean forecast tool. The agent calls it, gets point forecasts + uncertainty bands + a compact trend/seasonality summary, and writes the explanation and recommendation itself.

Quickstart (30 seconds)

uvx forecast-mcp        # runs over stdio for local agents

Add to your Claude Desktop / Claude Code / Cursor config:

{
  "mcpServers": {
    "forecast": { "command": "uvx", "args": ["forecast-mcp"] }
  }
}

Then ask your agent: "Forecast the next 6 months from this revenue data and tell me what to expect."

Enable the foundation model

pip install "forecast-mcp[timesfm]"

The server auto-detects TimesFM and uses it; otherwise it falls back to the statistical baseline. Both backends always return a result — no configuration needed.

Tools

Tool	What it does
forecast	Forecast a single series with optional uncertainty bands.
list_backends	Report which engine is active (timesfm / baseline).
backtest	Hold out the last N points and compare TimesFM vs baseline performance (MAE/sMAPE).

Documentation

Full docs at ramdhavepreetam.github.io/forecast-mcp:

• Getting Started — installation and first forecast
• Client Setup — Claude Desktop, Claude Code, Cursor configs
• Tool Reference — full parameter docs
• Cookbook — SaaS MRR, e-commerce demand, traffic, cloud spend
• How It Works — the math and model

License

Apache-2.0