The market catalog, built as one queryable system.
ISO publications, federal records, grid infrastructure, and weather — reconciled into a single catalog of 95 datasets. Every table keeps its native meaning and its own clock, and every number stays attached to the source it came from.
Every dataset, browsable and source-attached.
Filter by category, source, and cadence; open a dataset to read its coverage window, publication vintage, and the record it came from.
The warehouse behind the catalog.
Behind the catalog sits a warehouse of 32.4B stored rows, summed across the 94 of 95 catalog tables that report a count. Five-minute prices, per-resource dispatch, and hourly forecast vintages accumulate fast: ERCOT LMP by Bus alone holds 16.1B rows.
ISO coverage runs deepest in ERCOT Fuel Mix Report (2007), PJM Day-Ahead Transmission Constraints (2010), and ERCOT Hourly Historical System Load (2013). The deepest EIA record, Plant Monthly Generation & Fuel, starts in 2001. Every dataset carries its own start date, read from the data rather than asserted.
Rows in the catalog
32.4Bstored
94 tables counted
Largest single feed
16.1Brows
ERCOT · LMP by Bus
Datasets cataloged
95
Publishing sources
9
Domain categories
6
5-minute live feeds
32
Disclosure-lag feeds
2
Deepest history
2000onward
Open-Meteo · Weather History Grid (ERA5)
The catalog is free to read.
Every dataset is free to browse — schema, coverage, cadence, source link, and a live sample, no account required. An account costs $0 and takes seconds — no card, ever — and it carries the live maps for ERCOT, NYISO, CAISO, and PJM, asset profiles, saved work, and the grid agent at 40 questions an hour. Direct SQL, programmatic access, and private-data workspaces are priced separately.
Dataset, source, and category totals are computed from the catalog registry. Row counts and start dates are read from the warehouse hourly — stored rows as the engine reports them, before replacing-merge collapsing. A feed the warehouse does not report, or one indexed by a project milestone date rather than a publication clock, renders as —.
What the catalog is made of.
Every dataset, sourcedBy domain
6 categories- 0129Markets
- 0227Generation
- 0314Load
- 0414Infrastructure
- 058Transmission
- 063Weather
By source
9 feeds- 37NYISONew York Independent System Operator
- 27ERCOTElectric Reliability Council of Texas
- 9CAISOCalifornia Independent System Operator
- 9PJMPJM Interconnection
- 5EIAU.S. Energy Information Administration
- 3HIFLDHomeland Infrastructure Foundation-Level Data
- 2ODINOak Ridge National Laboratory outage feed
- 2Open-MeteoERA5 reanalysis and forecast weather
- 1NYSERDA IEDR
Three ways to reach the data.
The same catalog answers a point-and-click browse, a plain-English question, and a typed programmatic call. Query it however you work; every path carries source and cadence with the result.
The live catalog
Browse every dataset with its coverage, cadence, source link, and a live preview of the newest rows — then open any table on the live grid.
- Query
- Point-and-click: filter by market and category, read the schema.
- Export
- Each dataset page carries its source of record and freshness.
The grid agent
Ask in plain English. The agent searches the catalog, calls typed grid tools, and answers with the source and as-of stamp attached to every number.
- Query
- Natural language over the same tables, no SQL required.
- Export
- Answers move the map and keep their evidence trail.
MCP & tools
Point a compatible client at a scoped Ask the Grid server. Discover datasets, call typed queries, and pull series into your own tools and reports.
- Query
- Programmatic: search_data, describe_data, query_data.
- Export
- Structured results built for downstream export and automation.
From publication to evidence
One flow feeds the map, radar, agent, and MCP — and every path out carries the source and as-of with it.
A sample of what's inside.
See all datasets8datasets across the catalog’s domains and markets, with the rows each one holds and the year its coverage starts — both read from the warehouse, not estimated. Rows are stored rows, as the engine reports them before replacing-merge collapsing; a feed indexed by a project milestone date rather than a publication clock has no coverage start, and shows —. Each opens in the live catalog with its full schema, coverage calendar, and source of record.
Honest about time and gaps.
The credibility wedge is discipline about provenance. Freshness is read per feed, disclosure lag is disclosed, and a missing source is drawn as missing.
One clock per feed.
Five-minute real-time prices and annual infrastructure inventories never share a clock. Each dataset keeps its native cadence and publication vintage, so freshness is read per feed — not asserted for the whole catalog.
Disclosure lag is shown, not hidden.
Feeds released on a regulatory delay — like ERCOT's 60-day SCED disclosures (2 datasets) — render the trailing window as an expected lag, not a gap. What is not published yet is marked as such.
A missing feed reads as unavailable.
When a source has not published, the surface renders an empty or coverage state. Nothing is interpolated, back-filled, or fabricated to fill the space.
Your private data, alongside the public catalog.
A utility brings its private grid data and internal models; a trading desk brings CEII and positions. They meet the public catalog at an explicit boundary — analyzed together, with provenance kept legible on both sides.
- Utility
- A utility joins its private grid data and internal models to the public market surface — analyzed together, never commingled.
- Trader
- A desk brings CEII — Critical Energy Infrastructure Information — and positions into the same workspace, scoped to your tenant alone.
- Ownership
- The seam stays legible: every answer shows which tools read public records and which crossed the boundary. Provenance never blurs.
Open the catalog. Follow any number to its source.
Browse 95 datasets across 9 sources in the UI, ask the agent in plain English, or pull it into your own tools over MCP. No card, no demo call.