For developers

Accessibility in your pipeline, not a PDF in a drawer.

Everything the dashboard does is an API. Gate your builds on regressions, turn failing criteria into tickets, and wire the agent into your own product — with a typed SDK, React hooks, a CLI, webhooks, and MCP.

API reference Quickstart

Findings become tickets

Failing criteria open as GitHub issues.

Connect a repo and each failing WCAG criterion becomes an issue — deduped by criterion, auto-closed when a later scan finds it fixed, reopened on regression. Your engineers work accessibility in the tool they already live in.

Your token is verified, then stored in KV — never in the database.

# .github/workflows/a11y.yml
- uses: wholisphere/scan-action@v1
  with:
    urls: https://staging.example.com
    fail-on-regression: true
    github-issues: true

The developer surface

TypeScript SDK: A typed client for the whole API — scans, findings, VPATs — with built-in webhook signature verification. Published on npm under @wholisphere.ai.
React hooks: A provider plus hooks (useDescribe, useReadPage, useSummarize, useSimplify, useVpatGenerate, useFindings…) to wire the agent and findings into your own UI.
CLI: Generate and render VPATs from a stored scan in any format (JSON, OpenACR, HTML, Markdown, EARL), and lint reports in CI. Self-contained, no workspace needed.
MCP server: Expose accessibility capabilities to AI tools like Claude Desktop and Cursor: describe an image, simplify text, summarize a page, run a voice action.
Signed webhooks: Subscribe to scan.completed, scan.failed, scan.regressed, vpat.published and more — HMAC-SHA256 signed, retried with backoff, dead-lettered.
GitHub Action: Scan on every push or PR, post the results as a PR comment that updates in place, and fail the build when conformance regresses.

Model transparency

Which models we use.

The agent runs on a provider-agnostic router: Claude for vision and complex reasoning, Gemini for routine tasks, with automatic failover. Bring-your-own-key is supported in the extension for a subset of capabilities (describe, summarize, explain, translate, define) using your own Anthropic or Google key.

We’d rather tell you exactly what runs where than say "any model." Additional providers are on the roadmap, not shipped.

Read the docs, request a key.

The free tier includes API access. We're invite-only while in beta — request access and we'll email you when a spot opens.

Open the docs Request access