F.01
Plain-English insight
Every violation comes with a one-line user-impact line — what your blind, low-vision, motor-impaired, or cognitively-impaired user actually experiences. Not 'aria-required-attr failed.'
Audit. Insight.
Fix. Ship.
Most a11y tools stop at the report. Loop11y hands the report, the plain-English insight, and the patch — straight to your AI agent. Ten seconds in. Merged PR out.
▸ Run an audit
· No signup· ~10s / page· WCAG 2.2 · EN 301 549 · Section 508
Chapter 01 The Loop
Audit is step one of five.
Every accessibility tool stops at the report. We close the loop end-to-end so accessibility becomes a merged PR, not a backlog ticket.
01
Audit
Visible-first WCAG 2.2 against any URL, repo, or local file.
02
Explain
Plain-English user-impact for each violation — no rule-ID soup.
03
Rank
Contrast / font / theme / tap-target before semantic-only rules.
04
Patch
Mechanical fixes auto-apply. Judgement calls go to your agent.
05
Verify
Re-audit, side-by-side diff, score delta. Ship the PR.
Chapter 02 Features
Not a report. A fix loop.
Detection is a rounding error. The hard part is turning violations into shipped code without inventing what a tool cannot decide. Loop11y owns the entire arc — audit, explain, rank, patch, verify — and hands each step to your agent.
F.02
Visible-first ranking
Contrast, theme, font, and tap-target failures are what users notice. Ranked above semantic-only rules so your fix list mirrors perceived quality, not axe rule IDs.
F.03
Patch-ready for your agent
Each issue ships with a before/after code snippet your AI agent can apply directly. Claude Code, Cursor, Cline, Copilot — all speak Loop11y via MCP. Audit → diff → commit.
F.04
Mechanical auto-fix
Lang attribute, button names, decorative alt-empty, ARIA required-attr — all patched automatically. Judgement calls (alt copy, contrast colour) go to assisted mode. We never invent meaning.
F.05
Repo-wide source mapping
audit:repo scans React, Vue, Svelte, Angular, and plain HTML; maps every violation back to the exact source file with a confidence score so your agent can diff in one shot.
F.06
Verified, then shipped
Run audit → apply diff → re-audit. The loop closes itself. Gate PRs at a score threshold; merge confidently knowing the fix didn't regress somewhere else.
Chapter 03 The Report
Generated locally. Shared anywhere.
A single self-contained HTML file. No JS runtime, no SaaS storage. The same report you see below is what you generate for your site.
Chapter 04 Surfaces
Five surfaces. One open-source engine.
Pick the surface that matches where you work. All five run the same axe-core 4.10 audit pipeline. No vendor lock-in, no per-seat pricing.
01
Agent SkillOne-line install for Claude Code, Cursor, Cline, Codex, Aider, and 50+ other agents. Guided prompting on top of the MCP server.
$ npx skills add tayyabataimur/loop11y-skill -g -a <your-agent>02
CLIRuns locally — no SaaS, no telemetry. JSON, Markdown, SARIF, or single-file HTML output. npm package.
$ npx -y loop11y audit https://your-site.com --html --output report.html03
MCP serverstdio + streamable-HTTP transport. Works with Claude Desktop, Claude Code, Cursor, Cline, and any MCP-compatible agent.
$ npx -y loop11y # starts MCP server on stdio04
GitHub ActionGate PRs on accessibility score. Posts a Markdown report as a PR comment and fails the check below your threshold.
# .github/workflows/a11y.yml
uses: tayyabataimur/loop11y@v0.1.0
with: { url: https://preview.app, fail-under: 90 }05
HTTP APIOpenAPI-spec backend. Build a custom GPT, chat bot, or in-house dashboard. Self-host via Docker or Fly.
$ curl -X POST localhost:3000/api/evaluate -d '{ "url": "https://your-site.com" }'Chapter 05 Agent
Hand the report to your AI agent.
Loop11y exposes its audit, remediation, and verify tools over MCP. Your agent calls them like any other function. Paste a prompt; ship a PR.
Prompt · Claude Code / Cursor / Cline
▸ Audit https://staging.acme.com, rank the top 5 visible issues, and open a PR that fixes the auto-fixable ones. Re-audit and confirm the score moved.→ Agent calls
evaluate, then remediate, then verify. You review the diff.CI gate · GitHub Action
# Block merge if accessibility regresses
uses: tayyabataimur/loop11y@v0.1.0
with: { url: preview-url, fail-under: 90 }→ PR comment with the visual report. Status fails below threshold.
Programmatic · Harness SDK
› const r = await loop11y.evaluate({ url });
› if (r.score < 90) await loop11y.remediate({ mode: "fix" });→ Drop into your in-house dashboard, on-call bot, or release pipeline.
Chapter 06 Stakes
Reach the users you've been losing.
Accessibility isn't a compliance checkbox — it's the difference between a customer who can buy and one who bounces. The European Accessibility Act took effect 28 June 2025 and made that math explicit. Fix the loop and the addressable market grows.
+25%
addressable market unlocked by reaching disabled users (Eurostat 1-in-4).
71%
of disabled users abandon a site they cannot use (WebAIM 2024).
€20k+ /breach
administrative fines under EAA in DE and FR. Class action exposure on top.
96.3%
of the world's top 1M homepages had a WCAG failure last audit (WebAIM Million).
Audit. Then ship the fix.
One paste runs the audit. One MCP call lets your agent open the PR. One re-run verifies the score moved. That's the loop.