Intent
Describe what people need to understand and do before asking an AI assistant to build.
Open infrastructure for accessible software
Accessibility starts with what people need, in the brief and the prompt. ShapeA11y carries that intent into design tools, code, AI assistants and checks.
Give AI assistants accessibility rules before they write UI, then check what they build as part of your coding workflow. Start with the CLI, or choose the packages you need.
Get the CLI on npmnpm install --save-dev @shapea11y/cli
npx playwright install chromium
The CLI checks accessible names, heading structure, accessibility trees, keyboard focus and APCA contrast. See the README for usage.
Add rules and skills to a repository with npm create @shapea11y.
Connect agents to the inspection tools through MCP.
Inspect an element’s accessible name, role and state.
Read a compact accessibility tree in Markdown or JSON.
Inspect the page’s heading outline.
Check real Tab order and missing focus indicators.
Read virtual screen-reader output. This is not NVDA or JAWS testing.
Measure APCA Lc contrast for a text/background pair. This is not a WCAG conformance result.
Give agents persistent rules for accessible UI.
Add skills for authoring, reviewing and hearing interfaces.
Shape intent. Shape interfaces. Shape evidence.
Describe what people need to understand and do before asking an AI assistant to build.
Ask for semantic HTML, clear names and keyboard access from the outset.
Check whether people can complete their task with choice and independence.
Bring automated checks into the coding workflow to find barriers early, before someone has to report them. Combine those checks with human review and assistive-technology testing. Record what was tested, what was found and what remains uncertain.
Running a business? Start with the tasks your customers depend on. Read our practical guide to customer access.
Illustrative example
intent Payment confirmation
semantics An accessible name and ordered actions
verification Check stated requirements
Help shape requirements and review real experiences. Your expertise should influence decisions, with time and resources for your participation.
Behind every control is a person trying to do something. Ask AI assistants to suggest accessible patterns early, and test those choices with people.
Make comparable access, choice and independence part of delivery. Fund ongoing review, act on reported barriers and show where participation has improved.
An open letter
A software update should not take away someone’s access. We’re calling on AI tool creators, business leaders, and engineering and design teams to make that a shared responsibility.
Read and sign the open letterSupporting the letter does not require using ShapeA11y’s tools.
Carry reviewed accessibility requirements into redesigns and updates. Check what changes before release.
Share what was tested, what was found and what remains uncertain.
Resource their participation in decisions and review. Keep responsibility with people.