Python · Node.js · MCP

WCAG compliance, with receipts.

Six tools. One mission: make accessibility testing verifiable, automated, and hard to ignore.

Lint & gate

# lint CLI output for accessible patterns pip install a11y-lint a11y-ci a11y-lint scan . --artifact-dir .a11y_artifacts a11y-ci gate --artifact-dir .a11y_artifacts

Scan HTML

# scan HTML with cryptographic provenance npm install -g @accessibility-suite/evidence-engine a11y-engine scan ./html --out ./results # get fix guidance for a finding pip install a11y-assist a11y-assist explain --json error.json --profile screen-reader

MCP config

// connect to Claude Desktop, Cursor, VS Code { "mcpServers": { "a11y": { "command": "npx", "args": ["-y", "@accessibility-suite/mcp-tools"] } } }

Evidence over assertions

Most tools stop at "you have 12 violations." This suite goes further: every finding is backed by a tamper-evident provenance record.

Cryptographic provenance

Every finding carries a prov-spec provenance record with SHA-256 integrity digests. Same input always produces identical output — no network calls, no randomness.

Low-vision-first output

All CLI tools follow the [OK]/[WARN]/[FAIL] + What/Why/Fix contract. Five accessibility profiles: low-vision, screen-reader, dyslexia, cognitive-load, and standard.

CI-native by design

Exit codes, scorecard JSON, and PR comments built for automated pipelines. Gate releases on WCAG regressions. Drop in as a GitHub Actions composite action.

The six tools

Detection through remediation — each tool hands off to the next.

Tool

What it does

Stack

a11y-lint

Scan CLI output for accessible error message patterns, produce scorecards

Python · PyPI

a11y-ci

CI gate with regression detection, allowlists, and PR comments

Python · PyPI · npm

a11y-assist

Fix guidance in five accessibility profiles from structured findings

Python · PyPI

a11y-evidence-engine

Headless HTML scanner with prov-spec provenance records

Node.js · npm

a11y-mcp-tools

MCP server — evidence capture and WCAG diagnosis for AI assistants

Node.js · npm

a11y-demo-site

Demo site with intentional violations for end-to-end testing

HTML

Get started

Lint CLI output

pip install a11y-lint
a11y-lint scan output.txt

Gate CI on regressions

pip install a11y-ci
a11y-lint scan . --artifact-dir .a11y_artifacts
a11y-ci gate --artifact-dir .a11y_artifacts

Scan HTML with provenance

npm install -g @accessibility-suite/evidence-engine
a11y-engine scan ./html --out ./results

MCP tools for AI assistants

npm install -g @accessibility-suite/mcp-tools
a11y evidence --target page.html --out evidence.json
a11y diagnose --bundle evidence.json --verify-provenance

How it works

Six tools that form a pipeline from detection through remediation.

1. Detect

a11y-lint scans CLI text for accessible error patterns. a11y-evidence-engine scans HTML and emits findings with prov-spec provenance chains.

2. Gate

a11y-ci consumes scorecards, enforces thresholds, detects regressions vs. the last run, and posts PR comments. Fails the build on new serious violations.

3. Fix

a11y-assist generates fix guidance tuned to five accessibility profiles. a11y-mcp-tools brings the whole pipeline into your AI assistant via MCP.