AccessPreflight
API docs Sign inTry it free

From source asset to reviewable evidence.

A scan is asynchronous, versioned, and safe to retry. The API separates machine findings from review work, then returns the evidence needed to repair or route the asset.

Try it freeBook a demo

A clean automated result remains no_detected_failures. It does not claim that the full asset is accessible.

LifecycleSubmit → inspect → route → rescan

One asynchronous lifecycle, normalized findings, explicit coverage, and unresolved human checks.

A lifecycle that fits real queues.

Long-running document and rendered-page checks should not hold an upload or release request open. The job contract stays explicit at every state.

  1. 1

    Create the scan

    Submit a URL, HTML, PDF, or DOCX with a project, environment, and one or more pinned profiles.

  2. 2

    Accept the job

    Receive 202 Accepted immediately. An idempotency key protects retries from creating duplicate scans.

  3. 3

    Run source-aware checks

    The engine set inspects the format, normalizes findings, records coverage, and creates the human-review queue.

  4. 4

    Receive the outcome

    Poll the scan or consume a signed webhook when the result and configured reports are ready.

  5. 5

    Route the decision

    Compare with a baseline, apply a quality gate, repair the source, and rescan under the same profile version.

The profile is part of the request.

Pinning the profile makes a result reproducible. A later standards update creates a new profile version instead of silently changing historical evidence.

Browse available policy profiles
POST /v1/scans

{
  "asset": { "type": "url",
             "url": "https://example.org/checkout" },
  "profiles": ["wcag-2.2-aa-web@1.2.0"],
  "options": { "compare_to_baseline": "scn_…" }
}

Three outcome lanes, kept deliberately separate.

A machine observation, a contextual risk, and a human decision do not carry the same certainty. The API model prevents them from collapsing into one score.

Deterministic finding

The tested condition failed and includes reproducible evidence.

automation: full

Heuristic risk

The signal deserves attention, but context can change the conclusion.

automation: partial

Human review

The question requires meaning, task context, or assistive-technology experience.

decision: required

Run the first scan where it can change the outcome.

Connect before publication, keep the profile pinned, and make the next repair measurable.

Check a website freeCreate an account

Automated preflight is not an accessibility certification or legal determination.