Getting Started
This guide walks you through installing the Accessibility Suite tools and running your first accessibility scan. By the end, you will have a scorecard, a CI gate result, and fix guidance — all in under a minute.
Prerequisites
Section titled “Prerequisites”- Python 3.10+ for a11y-lint, a11y-ci, and a11y-assist
- Node.js 18+ for a11y-evidence-engine and a11y-mcp-tools
- pip and npm available on your PATH
Option A: Install from package registries
Section titled “Option A: Install from package registries”Install whichever tools you need. You do not have to install the entire suite.
Lint CLI output for accessible patterns
Section titled “Lint CLI output for accessible patterns”pip install a11y-linta11y-lint scan output.txtThis scans a file (or directory) for accessible error message patterns and produces a scorecard at .a11y_artifacts/current.scorecard.json.
Gate your CI on regressions
Section titled “Gate your CI on regressions”pip install a11y-cia11y-lint scan . --artifact-dir .a11y_artifactsa11y-ci gate --artifact-dir .a11y_artifactsThe gate command reads the scorecard, compares against any baseline or allowlist, and returns a pass/fail exit code.
Scan HTML with cryptographic provenance
Section titled “Scan HTML with cryptographic provenance”npm install -g @accessibility-suite/evidence-enginea11y-engine scan ./html --out ./resultsThe evidence engine scans HTML files and emits findings with prov-spec provenance records — each finding carries a SHA-256 integrity digest.
Get fix guidance for a finding
Section titled “Get fix guidance for a finding”pip install a11y-assista11y-assist explain --json error.json --profile screen-readerThe assist tool generates remediation guidance tuned to one of five accessibility profiles: low-vision, screen-reader, dyslexia, cognitive-load, or standard.
Capture evidence and diagnose via MCP
Section titled “Capture evidence and diagnose via MCP”npm install -g @accessibility-suite/mcp-toolsa11y evidence --target page.html --dom-snapshot --out evidence.jsona11y diagnose --bundle evidence.json --verify-provenance --fixOption B: Install from source (developer mode)
Section titled “Option B: Install from source (developer mode)”If you want to contribute or modify the tools, clone the monorepo and use editable installs:
git clone https://github.com/mcp-tool-shop-org/accessibility-suite.gitcd accessibility-suite
# Install Python tools as editablepip install -e src/a11y-lintpip install -e src/a11y-cipip install -e src/a11y-assist
# Install Node toolscd src/a11y-evidence-engine && npm install && cd ../..cd src/a11y-mcp-tools && npm install && cd ../..Your first end-to-end run
Section titled “Your first end-to-end run”The demo site provides a ready-made example with intentional accessibility violations. This is the fastest way to see all the tools working together:
cd examples/a11y-demo-site./scripts/a11y.shThis script runs the full pipeline: scan, gate, evidence capture, and comment generation. Inspect the .a11y_artifacts/ directory afterward to see all the outputs.
What to read next
Section titled “What to read next”- Architecture explains how the tools connect and how data flows between them
- Tools provides a detailed guide to each tool
- CI Integration shows how to drop the gate into GitHub Actions