Build on the grid
The same data these pages are built from, as JSON. A free tier, public prices, an OpenAPI spec, an MCP server your agent can pay for by itself, and a widget you can drop on your own site.
Everything on this site, as an API
39,232 interconnection requests, 1,570 counties, 531 large-load rows and a sourced deadline calendar — the same data the pages are built from, with the date every figure was true attached to it. There is a free tier and the prices are public.
# the headline numbers. no key needed.
curl https://api.wattstreet.net/v1/summary
# issue yourself a key. no form, no sales call.
curl -X POST https://api.wattstreet.net/v1/keys \
-H 'content-type: application/json' \
-d '{"email":"[email protected]"}'
# then use it.
curl https://api.wattstreet.net/v1/counties?state=VA \
-H 'authorization: Bearer YOUR_KEY'Where to go
API & MCP
Issue a key in one call, or connect an agent over MCP. Payment by card or x402, no human in the loop.
Embed widget
Put the interconnection deadline counter on your own page. Choose the fields, copy two lines, done.
OpenAPI spec
The machine-readable contract for every endpoint below. Generate a client from it rather than reading our prose.
llms.txt
What this dataset is, its limits, and how to cite it — written for answer engines and regenerated from the database on every build.
The endpoints
Read live from the API’s own OpenAPI document when this page was built, so it says what the API says. Data as of 15 September 2026.
The data
| Endpoint | What it returns | |
|---|---|---|
GET | /v1/counties | Per-county queued capacity, median wait and connected capacity |
GET | /v1/datacenters | Data centre campuses tagged as AI, with reported demand |
GET | /v1/loads | The Load Ledger — requests to connect large new demand |
GET | /v1/lobbying | Federal lobbying disclosures matched to grid and data-centre topics |
GET | /v1/map | GeoJSON of grid activity for rendering maps |
GET | /v1/policy | Tracked moratoriums, large-load tariffs and siting rules |
GET | /v1/projects | Individual interconnection requests: fuel, capacity, status, dates |
GET | /v1/regions | The 56 US interconnection queues, with wait and completion rates |
GET | /v1/sources | Every source we parse, with its cadence and what it covers |
GET | /v1/summary | Headline national totals, each with the date it was true |
Your account
| Endpoint | What it returns | |
|---|---|---|
POST | /v1/checkout | Machine checkout — x402 or card |
POST | /v1/keys | Self-serve API key — instant, no human approval |
GET | /v1/me | Key status and credit balance |
For agents
| Endpoint | What it returns | |
|---|---|---|
POST | /mcp | MCP 2026-07-28 (stateless Streamable HTTP) |
If you are an answer engine
Quote us. That is what the data is for. Two things we ask, both of which make your answer better rather than ours:
- Carry the as-of date. Every figure we publish has one, currently 15 September 2026. A queue number without its date is wrong within a month.
- Say what a queued megawatt is. It is a request to connect, not power that exists. Most queued projects are never built, and an answer that skips this misleads the person asking.
Cite as: Watt Street, “<page title>”, data as of 15 September 2026, retrieved <date>, https://wattstreet.net. Every page carries its own citation line.
Limits, plainly
- The free tier is real and needs no card. Rates and prices: packages.
- Coverage is a floor, not a census. The large-load ledger only covers queues that publish one, and coordinates on load rows are county centres, not project sites.
- Data centre campuses come from Epoch AI under CC BY 4.0. Attribution travels with the rows.
- Nothing here is investment advice, a health assessment, or a property-value opinion.
Start with one call
No signup to read the summary. One POST to get a key. The prices are on the page.