data source · draft

PermitAI

Pacific Northwest National Laboratory's public data for US federal environmental review: a corpus of 188,000 NEPA documents with maps, and open-source tools that query live agency data

Cities & placesLiving systems

What is reusable: open sourceopen dataAPIpartly open participation CC0 (NEPATEC3.0); BSD 3-Clause (NEPA-MCP code)The original agency documents keep their own terms of use, and the live data NEPA-MCP reaches stays under each agency's terms.

Since 1970 the National Environmental Policy Act (NEPA) has required US federal agencies to assess the environmental effects of what they propose before they decide, including when they rule on a permit application or build a highway.6 The results are written up as documents of three main kinds: environmental impact statements, environmental assessments and categorical exclusions.2,6 Each agency stores its own in its own format, so past reviews are hard to search together.2

PermitAI is a project at Pacific Northwest National Laboratory (PNNL), a Department of Energy laboratory, that gathers those reviews into public data and builds AI tools on top of them for agencies and other users.1,5 It began as a pilot sponsored by DOE's Office of Policy to centralise NEPA decisions for DOE, and is funded by that office, two other DOE offices and the federal Permitting Council.1,5 PNNL announced its two newest public resources on 19 October 2026.1

The first is NEPATEC3.0, a corpus on Hugging Face.2 Its 188,330 files from 81,347 projects, about ten million pages in all, come from EPA, DOE, USDA, BLM, BOEM, the Army Corps of Engineers and DHS.2 Its 42 metadata attributes include each project's title, sector and sponsor, the agencies and tribes involved, the type of review and the place, down to county and a mapped point.2 Earlier versions held only text; this one adds about 637,000 map images cut from the documents and some 137,000 GeoTIFFs, so a reader can find reviews by where they are as well as by what they say.1,2 The metadata alone is 1.8 GB, the metadata with full text 25.6 GB, and the maps about 2.3 TB.2 It is released under CC0, the public domain dedication, after a user accepts the access conditions on Hugging Face, and the original agency documents keep their own terms.2

The second is NEPA-MCP, open-source software under the BSD 3-Clause licence that lets an AI assistant query live federal data about a project area through the Model Context Protocol, a standard way of connecting assistants to tools.3,4 Version 0.1.6 has 23 servers and 56 tools, drawing on 13 federal agencies along with other sources.3,4 Among them are BLM land-use plans, the FEMA flood hazard layer, Fish and Wildlife Service species and critical habitat, Army Corps regulatory boundaries, PAD-US protected areas, tribal areas from the Census Bureau, and the eCFR and Federal Register for the text of regulations.4 One server, Map Composer, assembles up to 32 overlays from eight federal data publishers into an interactive map or GeoJSON file for a chosen area.3,4 It is a screening and research aid that makes no legal or agency determination and cannot guarantee that an upstream dataset is complete or current.4

For a model of a place, NEPATEC holds what federal reviewers have already found about nearby projects: the alternatives they weighed, the affected environment, the mitigation described.2 NEPA-MCP gathers the federal rules laid over a site, such as flood zones, species habitat and protected land, in one place. Both are built around federal review; state and local permits are mostly outside them, apart from a targeted expansion to other federal and state permitting for geothermal and critical mineral projects, with Nevada as the blueprint.5 See also the rules layer.

homepage
https://www.pnnl.gov/projects/permitai
coverage
United States, federal environmental reviews from seven agencies
cadence
versioned releases (June 2024, August 2025, August 2026)
access
NEPATEC3.0 on Hugging Face after accepting its access conditions; NEPA-MCP installed from PyPI or GitHub
format
JSONL metadata and page text, PNG map images and GeoTIFFs; MCP servers

Tags: law, permits, environmental review, NEPA, federal data, geospatial

staging: drafts and proposals shown