Energi Data Service MCP
Energi Data Service MCP
Read-only remote MCP server for Energinet's public Energi Data Service.
Tools
search_datasets— search dataset metadata.get_dataset— query any named dataset with dates, columns and filters.get_co2_emissions— retrieve 5-minute CO2 intensity for DK1 or DK2.
No Energinet API key is required.
Run locally
npm install
npm startThe MCP endpoint is http://localhost:3000/mcp.
Deploy
ChatGPT requires a remote HTTPS MCP endpoint. Deploy this repository on a
Node-compatible host such as Render, Railway, Fly.io or Google Cloud Run.
render.yaml and Dockerfile are included.
After deployment, the endpoint is:
https://YOUR-HOST/mcpConnect in ChatGPT
Enable Developer mode in ChatGPT under Settings → Apps → Advanced settings. Workspace permissions may require an admin/owner.
Choose Settings → Apps → Create.
Enter the deployed HTTPS endpoint ending in
/mcp.Choose No authentication.
Select Scan tools, review the three read-only tools, and create the app.
Start a new conversation and enable the app from the tools menu.
Connect in Claude
Claude Pro or Max
Open Claude and go to Customize → Connectors.
Select + → Add custom connector.
Enter the deployed HTTPS endpoint ending in /mcp.
Leave OAuth credentials empty because the server uses no authentication.
Select Add to save the connector.
Start a new conversation.
Open the tools or connectors menu and enable the custom connector.
Ask Claude to use one of the available energy-data tools.
Example:
Fetch the CO₂ forecast and day-ahead electricity prices for DK1 and identify the best time to consume 1,000 kWh.
Claude Team or Enterprise
An Owner or Primary Owner must first add the connector:
Go to Organization settings → Connectors.
Select Add.
Hover over Custom and select Web.
Enter the deployed HTTPS endpoint ending in /mcp.
Leave the optional OAuth Client ID and Client Secret empty.
Select Add.
Example requests
"Find Energinet datasets about electricity prices."
"Get CO2 intensity for DK1 for the past 24 hours."
"Get records from DeclarationProduction for DK1 in January 2026."
Operational notes
The upstream API applies dataset-specific rate limits. Avoid polling faster than the source dataset's update frequency.
Generic queries are capped at 5,000 records per call.
The server does not store user data or upstream responses.