a11y-checker MCP
An accessibility scanner for agents and CI: renders a page in Chromium with axe-core or lints React, Vue and Angular source.
- 117 ★ on the repository
- Last commit
Minimal access
Why this level
- Launches Chromium and opens the URLs you give; the agent edits source only in the files you name.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/a11y-checkerDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- Launches Chromium and opens the URLs you give; the agent edits source only in the files you name.
This is third-party code. Review the repository files before installing.
What it does
a11y-checker checks accessibility two ways. The web engine opens a URL or HTML in Chromium and runs axe-core, so it works with any framework. The source engine lints code without a running app: React and Next.js, React Native, Vue, Astro and Angular templates. Every finding is a structure with `ruleId`, impact level, WCAG criterion and a fix hint. For agents there is an MCP server with `scan_page` and `explain_rule`, an `a11y-fix` skill and an `AGENTS.md` block. The agent scans, fixes and rescans only the files you name.
Who it is for. Frontend developers and QA engineers who want an agent to find and fix WCAG violations itself and CI to fail on serious ones.
Good fit when
- An agent finished UI edits and the result is checked against WCAG.
- CI gets a check with `npx a11y-checker src --fail-on serious`.
- An app built on Svelte or a template engine: scan the URL of the running version.
- After edits only changed files are checked with the `--changed` flag.
Not a fit when
- You need an on-device check of a native app: React Native has no simulator scan.
- You need a full audit with manual screen reader testing: automation finds only part of the problems.
- Node is older than 22: the package does not support it.
Example request
Check accessibility of `src/components`, fix the serious violations, then rescan and show the report.MCP
- Transport
- stdio
- Authentication
- not required
Limitations
Node 22 or newer is required. For web scans Chromium is downloaded separately with `npx playwright install chromium`. Svelte and template engines such as Handlebars and EJS have no static rules, so scan the URL. Automated checking covers only part of the WCAG criteria. Version 3 is a rewrite; the README describes migrating from version 2. React Native has only a static source check and no on-device scan.
How to disable. Remove the MCP block from the client config, run `npm uninstall a11y-checker` and delete `a11y-fix` from `.claude/skills` or `.agents/skills`. Delete the `AGENTS.md` block by hand.
FAQ
How do I connect the MCP server?
Add a stdio server with the command `npx -y a11y-checker mcp`. The top-level config key depends on the client: `mcpServers`, `servers` or another.
How do I install the skill?
Run `npx a11y-checker install-skill`: the skill lands in the project `.agents/skills` and `.claude/skills`. The `--global` flag writes to the home folder.
What does the source engine check?
React and Next.js (JS, TS, JSX, TSX), React Native, Vue SFC, Astro and Angular templates. For files without rules it prints a warning.
How do I set a CI threshold?
Use `--fail-on` with `minor`, `moderate`, `serious` or `critical`. Exit code 1 means findings at or above the threshold.
Is a browser needed to check source?
No. Chromium is needed only by the web engine: for URLs, HTML files and pages found in a directory scan. Static code linting does not use it.
Discussion
No questions or reviews yet.
Related
MCP serversThe uisight MCP server opens mobile and desktop versions of a site and reports contrast and button sizes as exact numbers.
Accessibility Agents
accessibility-agents
Accessibility checking agents for Claude Code, Copilot and Claude Desktop: WCAG 2.2 AA requirements as skills
Annotation toolbar for React pages: click an element to hand the agent its selector and context, plus skills for setup and an autonomous UI critique
Console, network, screenshots and Lighthouse audits from your open Chrome session to the agent via a DevTools extension
An MCP server plus a Chrome extension: pick a DOM element by Option-click and the agent gets its text, styles and selector
A CLI that sets up a canvas where an AI agent builds dashboards and reports as React artifacts, revised block by block
README in short
One package provides a CLI, a library, an MCP server, a skill and an `AGENTS.md` block. Exit codes and JSON output fit CI.