Voltcast MCP Server — European electricity prices, forecasts & carbon

Model Context Protocol server for Voltcast: day-ahead electricity prices for 47 European bidding zones (15-minute native resolution, history to 2015), probabilistic price forecasts whose accuracy is published daily at voltcast.com/accuracy, carbon intensity, and cheapest-window optimization.

Two ways to run it:

1. Hosted endpoint (recommended)

Code

https://voltcast.com/api/mcp

Streamable HTTP transport, protocol 2025-06-18. Server card: /.well-known/mcp.json.

2. Local stdio server (this repo)

bash

VOLTCAST_API_KEY=your_key npx voltcast-mcp

Or with Docker:

Terminal

docker build -t voltcast-mcp .
docker run -i -e VOLTCAST_API_KEY=your_key voltcast-mcp

The stdio server is a thin proxy onto the same REST API — identical tools, auth and rate limits. It starts and answers introspection without a key; list_zones works keyless too.

config.json

{
  "mcpServers": {
    "voltcast": {
      "command": "npx",
      "args": ["voltcast-mcp"],
      "env": { "VOLTCAST_API_KEY": "YOUR_API_KEY" }
    }
  }
}

Tools

Tool	Auth	What it does
`list_zones`	none	The zone registry (codes, names, currencies)
`get_prices`	key	Day-ahead prices for a zone/date at native market intervals
`get_forecast`	key	P50 on Home/Starter and P10/P90 on Pro+
`get_carbon`	key	Carbon intensity + green score from the live generation mix
`get_renewables`	key	Wind/solar forecasts (TSO + Voltcast model) vs realized
`cheapest_window`	Pro+ key	Cheapest contiguous window for a load in the next 48h

Auth

Authorization: Bearer <api key>. New accounts start with Home at €9/month after a 7-day card-required trial; existing Free accounts remain grandfathered. list_zones works without a key. The exact hosted-connector walkthrough is at voltcast.com/ai/local-electricity-price.

Client config

config.json

{
  "mcpServers": {
    "voltcast": {
      "url": "https://voltcast.com/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

Honesty, stated plainly

Every forecast the server returns is benchmarked daily against a naive baseline in public — losses included. Zones where the model loses are free to use under the guarantee. Data sources: ENTSO-E Transparency Platform, SMARD.de (CC BY 4.0), attributed in every response.