Ask the Grid
LiveMetiscast · probabilistic price forecast · ERCOT nodal

Metiscast Nodal

metis/ercot-nodal

Probabilistic price paths for individual ERCOT resource nodes.

Price forecasts at the generator- and load-level surface of the ERCOT market — the resource nodes where a specific unit actually settles, each with its published p10–p90 band.

How it works

A post-trained derivative of the Geotemporal Foundation Model, narrowed to ERCOT locational marginal price. It inherits the base model's joint attention over time and space, which is what lets a node with thin history borrow shape from the nodes physically and electrically around it. Output is a distribution per horizon rather than a point estimate, issued every 15 minutes at five-minute resolution across horizons from one hour to one week. Nodal, zonal, and hub are the same model addressed at three settlement geographies — they differ in what they are asked about, not in what was trained.

Inputs

  • ERCOT settlement-point prices — real-time (SCED) and day-ahead
  • System load, generation mix, and renewable output
  • Weather fields — forecast and observed
  • Grid geography — settlement-point coordinates and network topology

Outputs

  • A p50 price path with its p10 and p90 band, per horizon
  • Five-minute resolution across a 48-hour realized and forward window
  • An issuance timestamp, so a forecast can be scored against what it knew

Three ways to consume it

product · inference · MCP
01

In the product

Metiscast Nodal powers Ask the Grid directly — the live map, the radar, and the grid agent read it in place, so the same model you'd call is the one already reasoning over the grid on screen.

02

Developer inference service

A direct, typed inference API: request forecasts or embeddings by node and horizon and get structured results back. This is a plain inference endpoint — not MCP — for pipelines, notebooks, and services that want the model output without an agent in the loop.

See the developer surface
03

MCP

The same model outputs, exposed as tools your own agent can call. Point a compatible client at the Ask the Grid MCP server and the model becomes a tool in your agent's own loop.

Build with MCP

Intended use

  • Forecasting the price at a specific generator or load
  • Scoring a forecast vintage against what actually settled
  • Feeding dispatch, hedging, and position models that need a distribution rather than a point

Limitations

  • ERCOT only today; other markets follow as evaluation lands
  • Probabilistic, not a guarantee — the band is the forecast, not decoration
  • Bands widen with horizon; a one-week path is not a five-minute nowcast

Consume it in the product, over the inference API, or through MCP.

Every model reaches the same three surfaces. Start where you work and the model output comes back with its source and as-of attached.