♿ WCAG 2.2 Accessibility Auditor — Bulk axe-core

Give it a list of page URLs. Each page is rendered in headless Chrome, scanned with axe-core 4.10.0, and returned as a structured WCAG 2.2 report.

💵 Pricing

Event	Price
URL audited (one page rendered and successfully scanned)	$0.05 — about 20 audits per $1
Actor start	$0.005 per run, per GB of memory (minimum 1)

You are charged only for pages that were actually rendered and audited. A page that could not be loaded — blocked, timed out, DNS failure — is not charged and is not written to the dataset; it is listed in the run's OUTPUT record and in the run status message instead.

📊 What you get

One dataset row per audited page, with these exact fields:

Field	Type	Meaning
url	string	The page audited
audit_engine	string	axe-core 4.10.0
wcag_version	string	2.2
total_violations	int	Number of failing axe rules
total_passes	int	Number of rules that passed
total_incomplete	int	Rules axe could not decide (needs human review)
total_inapplicable	int	Rules with nothing on the page to test
severity_breakdown	object	{critical, serious, moderate, minor} rule counts
wcag_tag_coverage	object	Count of failing rules per WCAG tag (wcag2aa, wcag22aa, …)
accessibility_score	int	0–100, derived from violation severity × affected nodes
top_violations	array	Up to 25 failing rules, worst first — see below

Each entry in top_violations:

rule_id, impact, description, help, help_url, wcag_tags, affected_nodes, sample_selectors (up to 3 CSS selectors), sample_html (up to 3 snippets, 200 chars each), failure_summary.

Note on accessibility_score: it is this actor's own heuristic (critical −15, serious −8, moderate −3, minor −1, per affected node, capped at 20 nodes per rule, clamped to 0). It is a triage aid, not a certification, and it is not produced by axe-core. Nothing here is a legal conformance audit — axe-core catches roughly the automatable subset of WCAG; manual testing is still required for a conformance claim.

⚙️ Input

Field	Type	Notes
urls	array of strings	Required. One URL per line. Bare hostnames are upgraded to https://. Also accepts startUrls with {"url": …} objects.
wcagTags	array of strings	Optional. Leave empty for the WCAG 2.2 AA set: wcag2a, wcag2aa, wcag21aa, wcag22aa. Other values: wcag21a, best-practice.
timeoutMs	integer	Optional. Per-page load timeout, 5000–120000. Default 25000.
proxyCountry	string	Optional. Two-letter country used only for residential retries after a block.

An empty input ({}) does no work and charges nothing — the run succeeds with 0 rows and a message naming urls.

🔁 Blocked pages

A page that refuses the direct request (connection reset, timeout, TLS failure) is retried up to 3 more times through the Apify residential proxy, with a fresh session on each attempt. Only after all attempts fail is the page recorded as unaudited — and never charged.

🚦 How runs end

• Rows delivered — status message reports how many pages were audited and charged, and how many were not.
• Every requested page unreachable — the run FAILS with the first cause named. A blocked site is never reported as a clean audit.
• Browser could not start — the run FAILS; nothing is charged.
• Run time limit reached — the run stops cleanly before the platform timeout, keeps the pages already audited, and says "stopped early at the run time limit". Remaining URLs are listed in OUTPUT.skipped_urls.
• Nothing to do — missing urls succeeds with 0 rows and says which field to fill in.

⏰ Schedule it

Run on a schedule (0 8 * * 1) to watch accessibility regressions over time.

🤖 Use with AI agents

Point Claude, the OpenAI Agents SDK, an n8n flow, or any MCP-aware client at it.

Run WCAG 2.2 Accessibility Auditor on my input and return the structured results.

Agentic payments (x402): supports agentic payment via USDC.

🔗 Related tools

Lighthouse Bulk Checker · Page Speed Analyzer

---

A NexGenData utility actor. axe-core is MPL-2.0, © Deque Systems.