Prove access was delivered. Not just claimed.
Everyone else does one of three things: scan, “fix” with an overlay, or consult. Wholisphere closes the loop — scan your site, remediate it with an AI agent for real users at runtime, measure that the agent delivered access, and ship an audit-ready VPAT that cites the proof.
One line of code, or zero — we can scan for you. Currently in invite-only beta: request access and we'll email you when a spot opens. SOC 2 in progress.
In private beta with early design partners. Built with what we sell — powered by:
- Cloudflare Workers
- Claude
- Gemini
- Stripe
- Model Context Protocol
The loop nobody else can close.
Scanners find defects. Overlays claim to fix them. Consultants write reports. Only Wholisphere does all four steps — and measures the third, so the fourth tells the truth.
- Scan
Find the real defects
80+ evaluators across WCAG 2.2 — machine coverage of 50 of the 55 Level AA criteria. Vision-LLM and Playwright probes catch criteria rule-based scanners can't touch — keyboard traps, reflow, images-of-text, caption coverage.
- Remediate
The agent delivers access
Embedded on your site, the Wholisphere agent reads, describes, captions, simplifies, and voice-operates the page for the users who need it — coexisting with JAWS / NVDA / VoiceOver, never overriding them.
Try it live in the interactive demo below.
- Measure
Verify it actually worked
Our deepest evaluation tier runs the agent's own capabilities against your page and records, per criterion, whether it delivered access. Measured — never assumed. No vision key? It reports “unverified,” never a false pass.
- Attest
A VPAT that cites the proof
Every “Does Not Support” row pairs with a §1194.5 equivalent-facilitation entry backed by that measurement. Signed, hashed, chained into a tamper-evident audit log.
Competitors can't copy the last step — they have no remediation agent to verify.
Four products. One loop.
Scanning & Monitoring
Hosted scans (paste a URL, we run it), scheduled monitoring, regression alerts, and findings that auto-file as GitHub issues.
Explore scanning → AttestVPAT & Compliance
Auto-generated VPAT 2.5 INT mapped to Section 508 + EN 301 549. Review, approve, publish immutable versions — every claim backed by evidence.
See the VPAT product → RemediateAssistive Agent
A floating landmark on every page: read-aloud, describe images, live captions, focus mode, voice control. One line of code, or the browser extension.
Meet the agent → IntegrateDeveloper Platform
TypeScript SDK, React hooks, CLI, an MCP server, signed webhooks, and a GitHub Action that gates builds on accessibility regressions.
Build with us →Click Read this page below.
A live, interactive demo running our agent against a sample article. Read it aloud, hide clutter with focus mode, or overlay live captions. Nothing to install.
Demo runs entirely in your browser; nothing is sent to your tab without your action. Want it on a real site? Grab the embed snippet or install the browser extension.
Run accessibility as a service.
Onboard a client in one form. We scan on a schedule, you review and publish the VPAT, and the client watches a live status page — no account, no terminal, your brand.
See the agency workflow → For engineering teamsAccessibility in your pipeline.
A GitHub Action fails the build on regression. Failing criteria open as deduped GitHub issues that auto-close when fixed. Wire the rest through the SDK, webhooks, or MCP.
Read the developer docs →We are not an overlay.
| Overlay vendors | Wholisphere | |
|---|---|---|
| Claims | “100% WCAG compliant in seconds” | Equal access via mediated agent. Honest conformance reports. |
| Coexists with native AT? | No — fakes ARIA, breaks JAWS | Yes — the widget is a landmark, never overrides screen readers |
| Audit trail? | None | Every agent action logged + exportable as JSON |
| Proof the fix worked? | None — self-declared | Measured agent verification, per criterion |
| Lawsuits? | 200+ ADA lawsuits filed against sites using overlays | Architected with disability community input; hedged with real conformance reports |
Everything here is live.
Conformance badge
The latest scan's verdict, like a build badge. Re-fetched daily.
Live sample VPAT →
Real VPAT 2.5 INT rendered from the same pipeline customers use — including the Equivalent-Facilitation block a scanner without a live remediation agent can't produce.
Client status portal →
A tokenized, no-login page per product: current score, scan-history trend, and published VPATs.
/checkout
Automated scan on 2026-07-11 found 7 text nodes with a contrast ratio under the WCAG 2.2 AA
threshold. Lowest measured: 2.9:1 (.promo-banner .fine-print). Auto-closes when a later
scan finds this criterion passing; reopens on regression.
Ready to skip the refactor?
There's a real free tier — access is currently by invite. Request it and we usually reply within a day or two.