vpat vault Buy →

// accessibility case file · v1

Evidence in. Claims checked.

VPAT Vault turns accessibility evidence into reviewable VPAT documentation without asking a model to invent certainty.

No separate VPAT Vault API or backend is required; it runs in your existing Codex or Claude Code setup.

Buy now → Inspect the evidence trail
9 prompt recordslocal checkshuman sign-off required
Illustration of an accessibility audit under a magnifying glass
Exhibit A: automation finds signals. Reviewers decide what those signals mean.

// operating procedure

Scan → Systemize → Ship

One narrow workflow keeps source evidence attached to the language it supports—and stops clean scans from becoming unsupported “Supports” ratings.

  1. Scan

    Bring automated results, manual test notes, screenshots, and known product scope.

  2. Systemize

    Draft rows with nine prompts, then check labels, remark structure, and evidence boundaries locally.

  3. Ship

    Review every claim, assemble report JSON, and render fixed, escaped HTML.

// evidence, not vibes

Three checks between draft and delivery

The model drafts. Small deterministic tools guard the file boundaries. A human still owns the finding, rating, and final publication decision.

01 · language

Draft linter

Flags non-ITI labels, common defect language under “Supports,” thin WHAT/WHO/WHERE remarks, and Markdown leakage. Heuristic only.

02 · structure

Report renderer

Validates required fields, types, and ITI labels before escaping content into fixed, semantic HTML.

03 · delivery

Package builder

Builds from an explicit allowlist and records SHA-256 hashes in the buyer manifest.

// real worked example

One case file, run for real

The product's own prompts drafted this against a live scan of classiccottrell.ca, then a human reviewed the draft and prompt 9 plus render_vpat.py rendered it. It found a clean scan and refused to call that a "Supports" rating.

The example is not a certification and is not a delivered VPAT. It is a demonstration of the refusal path: 0 criteria rated, 1 incomplete finding flagged for manual review. Historical examples remain examples until their evidence is reproduced and reviewed.

View a VPAT example →

// product roadmap

Built in public, with evidence attached.

The first release keeps VPAT creation and scanning in one local workflow. Each next step earns its place by making evidence easier to collect, review, and carry into a defensible document.

Shipped · v1

Prompt workflow

Nine prompts, explicit output contracts, local validation, and a fixed HTML renderer for reviewable VPAT drafts.

Buy now →
Built · local

Scan evidence

The free Chrome extension runs axe-core locally and exports normalized findings plus manual-review items as JSON.

See the scanner →
Next · review

Attach context

Bring scan results, manual notes, screenshots, and product scope together before drafting criteria rows.

Planned workflow refinement
Later · launch gate

Package for teams

Prepare store-ready extension materials and a repeatable multi-page evidence session after privacy and product review.

Requires product-owner sign-off

// read before use

A clean scan is not a clean bill of health.

  • Automated results cannot establish a “Supports” rating on their own.
  • Every row requires human verification against source evidence and current ITI guidance.
  • VPAT Vault is not an audit, certification, legal opinion, or compliance guarantee.
  • Historical examples remain examples until their evidence is reproduced and reviewed.

// separate free tool · available for local install

Google Chrome extension

Capture the scan. Keep it local.

VPAT Vault Scanner is a Manifest V3 Google Chrome browser extension that runs axe-core 4.10.2 against the active tab and exports the findings as structured JSON, without a VPAT Vault service or account.

  • Permissions: scripting, downloads, sidePanel, storage, and host access to run on the pages you visit (needed so a rescan still works after you navigate — Chrome revokes narrower activeTab access on every page change). No tabs permission.
  • Scans still run entirely locally in the tab. Nothing is uploaded; the JSON leaves the device only when you choose to download it. Per-finding notes are stored locally via chrome.storage.local and never leave your device either.
  • Exports the same normalized shape (violations plus incomplete findings) your prompt 8 workflow expects.

Chrome Web Store listing coming later.

Chrome extension details →

Separate product. Separate release schedule. Automated findings still require manual testing and review.

// case ready for review

Document what the evidence can defend.

Nine prompts, explicit output contracts, local validation, a fixed renderer, and a buyer start guide.

Buy VPAT Vault →