What a Data Center Site Screen Misses, and How GridPermit Closes the Gap with Tariffs, the Local Record and an API

A site screen is a photograph. It shows a parcel, a utility, a county and a set of rules on the day someone ran it, and every one of those can change before a deal closes. A utility amends its large-load tariff. A county puts a pause on new water taps. A moratorium that was a single line on a draft agenda becomes an adopted ordinance. The parcel that screened clean on Monday is a different answer by the time the lawyers get involved, and nobody sent a notice.

GridPermit exists to shrink that gap. It started as a way to screen a site for a go or no-go answer. Over the last several weeks most of the work has gone into the part around the screen: keeping the underlying records current, showing how much to trust each one, and letting other people's software use all of it. This post walks through what that looks like today, including the places where it is still thin.

What the screen covers

A screen looks at a site across power, land, permitting and moratoriums, and returns a read that ends in a clear category, such as go, no-go, or pause pending confirmation. The pause category matters more than it sounds. When the evidence for a decision is weak, the honest answer is not a confident yes or no. It is a statement of what has to be confirmed first, and by whom.

Water appears next to the result and does not change the score. That is a deliberate choice. Water rules are local, uneven and often unwritten, and a number built on thin data would look more certain than it is. So the screen shows what is known about water near a site and leaves the weighing to the person reading it.

Every large-load tariff record also carries its own provenance: an evidence level, how many days old the review is, whether the utility's docket is watched every day, and the reasons the level is what it is. If a minimum bill or a collateral requirement is on the screen, you can see whether it came from a filed tariff or from something weaker. The goal is that nobody has to take the number on faith.

Watching the local record every day

Tariffs are only half of what changes a site. The other half happens in public meetings, on agendas that most screening tools never read. GridPermit now reads more than 240 agenda sources across 48 states every day: county boards, city councils, planning commissions, state agencies and utility boards, some through structured feeds like Legistar and the rest by reading the pages themselves. New lines that look like a data center item, a rezoning, a special use permit, a development agreement, a tax abatement, a large power request paired with a land-use action, are filed as leads for review.

This month the watch got broader on water. Many of the limits that actually stop a project never mention a data center. A town pauses new water taps or sewer connections. A utility caps its sewer or water capacity. A county considers a well-permit moratorium or a groundwater pumping cap. A city writes a rule aimed at large or industrial water users, or suspends will-serve letters for big new developments. The agenda watch now catches those lines on their own, and it deliberately ignores the routine items, such as watering days during a drought or a rate adjustment, that would bury the real ones.

A nightly job then does the slower work. Each night it goes through permitting records and water sources for a few states, upgrades what it can verify, searches the last sixty days of news and public records for local water and sewer limits in places the agenda list does not cover, and writes down what it found and what it could not. The log is part of the record, including the states where it found nothing.

Where the coverage is still thin

It would be easy to read the numbers above as complete coverage, and they are not. Several of the agenda sources are a city or county's general agenda page, so a water item shows up there as one line among many. A handful of states have few or no usable sources yet, because the official pages were blocked, moved or never published in a form a program can read. The curated list of adopted local water restrictions is still empty, and the nightly search is how it will fill. Each source also has to be read correctly, and some pages change format without warning.

The practical advice is simple. Treat a flagged item as a lead, not a finding, and treat silence from a thinly covered place as the absence of a source rather than the absence of a rule. When the stakes are real, call the clerk. The product is meant to tell you whom to call and what to ask, a lot faster than doing it from nothing.

Using the data in your own tools

Plenty of teams do not want another dashboard. They want the answer inside the spreadsheet, the CRM or the land-acquisition tool they already use, and that is what the API is for. You can screen a site on its own or send a batch, and download a batch as GeoJSON. You can pull tariff records, the history of a utility's docket pages with the lines that appeared and disappeared, and a log of changes to the tariff table over time.

A nearby endpoint lists gas pipelines and plants, generation, known data centers and, on plans that include it, the planned-projects watch list within up to 50 miles of a point, as JSON, CSV or GeoJSON. Watches let you subscribe to a site, and up to 25 can be created in a single call. Webhooks can filter by utility and post a readable message straight to Slack or Microsoft Teams, with a way to replay a past event if something on your end missed it.

One endpoint is aimed at a question that comes up in every negotiation. The benchmarks endpoint returns the median and range of contract term, ramp, minimum bill, collateral, exit fee and notice across the large-load tariffs that state each of those terms. Add the terms you were offered, and it tells you how many tariffs are smaller, equal or larger. It returns summary numbers only, so it does not use up your screening quota. It is a way to find out whether a utility's offer is ordinary or unusual before you answer it.

The API ships with an OpenAPI specification, a Postman collection, and JavaScript and Python clients. A public changelog lists every change that could affect an integration, with breaking changes marked. A usage endpoint shows your plan, your limits and the last thirty days of calls, and responses carry quota headers so your code can see how much is left.

What is open and what needs a key

Summary data stays open: the tariff list shows utility, state, status, threshold, source and review date without a key. The stated terms, evidence levels and docket flags need a key, and the CSV downloads need a free account. Paid keys accept the API terms once before they work, and new keys start at a reduced rate that ramps up over the first weeks. Those limits exist to keep the service steady for everyone, and the plan details are on the pricing page.

If you screen sites for a living, or build tools that do, the best test is your own list. Run the five parcels you know best and see whether the answers match what you already learned the hard way. And if there is a place or a kind of rule you wish this tracked and it does not, tell me. The coverage gets better fastest where someone points at a gap.