Skip to main content
ClearPathNEPA

Resources

Built on a published, interoperable contract

The same records you see in the UI are retrievable through a documented data-standard API, with provenance intact and ready for partner agencies and adjacent systems. Below: the leave-behind kit, a pilot offer, and answers to the questions agency buyers ask.

  • The data standard

    The data model is the federal NEPA & Permitting Data Standard: thirteen entities, canonical field names, and provenance on every record.

    Read the conformance notes

  • Interoperability

    One engine serves any agency’s CE set through a single OpenAPI contract. You integrate against a published interface, not a bespoke one.

    How multi-agency works

  • Architecture & readiness

    The production target (AWS GovCloud, FedRAMP Moderate, and NIST 800-53 control inheritance) alongside the human-in-the-loop and swappable-AI guarantees.

    Security & federal readiness

Leave-behind kit

Downloads

The booth collateral, hosted here so it doubles as Solutions Catalog material. The two PDFs are the concept-paper package as submitted in June 2026; the API is live in this build.

  • One-pager

    Concept collateral · June 2026 · PDF

    The booth leave-behind: ClearPath mapped to the Permitting Action Plan’s ten minimum functional requirements and four service delivery standards, on a single page.

    Download the one-pager

  • Architecture brief

    Concept collateral · June 2026 · PDF

    What a CISO asks for: the AWS GovCloud / FedRAMP Moderate / NIST 800-53 production target, the human-in-the-loop and swappable-AI guarantees, and the data-standard conformance story.

    Download the architecture brief

  • Live standards API

    OpenAPI 3 · served by this build

    The machine-readable data-standard contract this demonstration serves, with schemas, paths, and provenance fields. Browse it as an interactive console and fire a real request, or open the raw spec.

    Open the interactive API console

Live endpoints

Data-standard API

The published contract is exactly these four operations across three standard entities — the same set the interactive console renders. Project and process-instance routes take an ID from any case in the system.

  • GET/api/standard/process-instances/{id}

    Process instance

    The CE review process instance: decision elements, decision payload, and linked document, all standard-named.

  • POST/api/standard/process-instances/{id}

    Write-back

    Partner systems append comments and events to the same record. Every write is tagged with its source system and provenance.

  • GET/api/standard/projects/{id}

    Project

    A single project entity in standard shape, with full provenance (record_owner_agency, data_source_system, retrieved_timestamp…).

  • GET/api/standard/legal-structures

    Categorical Exclusion catalog

    The federal CE catalog as legal_structure entities: citation, issuing authority, and source for each of the 2,105 CEs.

Explore the interactive API console

Open the raw OpenAPI spec

Case artifacts

Downloadable records

Also live in this build, but sitting outside the data-standard contract: per-case downloads, keyed by any case ID. These are rendered artifacts, not standard entities, so they aren’t part of the OpenAPI console.

  • GET/api/cases/{id}/admin-record.json

    Administrative record

    The append-only event store for a case, serialized as machine-readable JSON.

  • GET/api/cases/{id}/determination.pdf

    Determination (PDF)

    The rendered, digital-first CE determination document for an approved case.

A concrete next step

90-day standards-conformance pilot

One bureau · one CE workflow · your CE catalog loaded from data · measured in days-to-determination.

What your agency provides

  • One CE workflow and the bureau’s CE catalog (or a pointer to your slice of CE Explorer).
  • A named reviewer and a few representative past determinations to set the baseline.
  • The extraordinary-circumstances checks and any agency-specific decision elements that matter to you.

What Nuvitek provides

  • Your catalog and decision elements loaded and configured from data, with no custom code.
  • A running instance on the NEPA & Permitting Data Standard, with the OpenAPI surface and append-only administrative record.
  • Days-to-determination measured against your baseline, with a written readout at day 90.

Start a pilot conversation

Questions agencies ask

Frequently asked

How do you keep the AI from inventing a Categorical Exclusion that doesn’t exist?

Structurally, it can’t. ClearPath retrieves real candidates from the 2,105-CE catalog first, and the model only ranks and explains those candidates. Every match it returns is joined back against the catalog by ID, and an invented ID is dropped rather than shown. So every citation on screen is a real catalog row. And no recommendation becomes a determination until a reviewer approves it, with that action written to the administrative record.

How is this different from CE Works?

We’re fans of it. CE Works proves CEQ believes in digitizing CE determinations, and the BLM Moab pilot proves agencies want it. ClearPath sits beside systems like it through the data standard rather than replacing them, and adds three things: AI-assisted screening against the full cross-agency catalog with citation and confidence, a data-standard OpenAPI surface with provenance on every entity (including write-back), and an append-only administrative record that distinguishes AI suggestions from human decisions.

Are you FedRAMP authorized, and how would an agency actually buy this?

The demonstration is deliberately not authorized; it runs locally on the real data standard so you can explore it without an account. The production architecture is documented for an AWS GovCloud landing zone on a FedRAMP Moderate baseline with NIST 800-53 control inheritance and Login.gov / PIV identity. That is a path toward an agency ATO, not a claim of one. On procurement, Nuvitek is an SBA 8(a) firm with active federal vehicles (GSA, IDIQs, BPAs), so a scoped pilot doesn’t require standing up a new vendor relationship.

Where does the data live? What about residency and multi-tenancy?

The demo uses a local Postgres. The production design is a single-agency, agency-owned GovCloud deployment: your data stays inside your boundary, not a shared multi-tenant store. The six provenance fields on every record (record_owner_agency, data_source_system, retrieved_timestamp, and the rest) are exactly what make sharing safe when you choose to exchange records across agencies. The standard, not co-tenancy, is what makes cross-agency interoperability work.

How does this satisfy AI governance, meaning M-25-21 and human-in-the-loop?

It’s enforced in code, not policy. Every AI output is a recommendation carrying confidence and a citation; a reviewer must approve it, and overriding a flag requires a typed rationale validated server-side. The decision payload records whether each result was AI-suggested or human-confirmed, and every AI call is logged with its model, latency, and prompt hash. That is OMB M-25-21 and agency human-in-the-loop direction, implemented as software. When the model or network is unavailable, a deterministic fallback produces the identical UI, and the record honestly tags which source produced each result.

Is it Section 508 accessible?

Yes. WCAG 2.1 AA holds throughout, built into the design system rather than retrofitted: semantic landmarks and headings, a skip link, visible focus indicators, AA contrast, reduced-motion support, and scrims behind text over photography. State is always conveyed by glyph and text, never by color alone.