Use cases & endpoints
EIA energy data for AI agents (inventories & production)
Energy markets run on a weekly and monthly release rhythm from the EIA: inventory levels, production estimates, consumption and trade. The weekly numbers move prices immediately; the monthly series revise the weekly picture afterwards, sometimes substantially.
That revision relationship is where naive datasets mislead. A weekly estimate later superseded by monthly data will, in a latest-value database, appear to have always been the revised figure — which makes the market's reaction to the original print look irrational rather than informed.
Hindcast keeps EIA series with each release as an observation and its publication timestamp attached, so the weekly print and its later revision both remain queryable. For an agent this means a question about what the inventory number was on a given morning returns the number traders saw, and a question about the eventual truth returns that separately, with no ambiguity about which is which.
hindcast — session
previewhindcast> inventory level as published that week, not as later revised
→ tool: eia_series { series:"...", as_of:"..." }
✓ the published value with its release timestamp
source: EIA release, revisions stored separately
What's covered
- Weekly and monthly energy series with release timestamps
- Inventories, production, consumption and trade flows
- Revisions stored as new observations, first prints retained
- Release calendars so scheduled publications are checkable
- Series joined to contract identifiers for price comparison
- As-of queries return what was published by that date
Join the waitlist — this page's dataset is already selected.
Questions
How are weekly and monthly series related?
They are kept as separate series with their own release histories. The monthly data often revises the weekly picture, and collapsing them would erase the distinction between an estimate and its correction.
Is this useful outside energy trading?
Yes — energy inventories are a widely used input to inflation and activity analysis, and having them point-in-time keeps that analysis consistent with the rest of a vintage-aware macro workflow.
Related
Join the waitlist. Vote the roadmap. First in, first served.