QualiBooth
English

New in Agora · MCP server

Your AI coding agent, now an accessibility expert.

Agora’s built-in MCP server gives Claude Code, Cursor, GitHub Copilot and any other MCP-capable agent a real accessibility testing tool. Your agent scans the page it just built, sees exactly what fails and why, fixes the code and scans again to prove it - before the pull request even exists.

Start your free trial Book a demo
Runs on your machine
Local-only server, local scans
One command
to connect your coding agent

Animated example of a coding session: a developer asks their AI agent to check localhost:3000/checkout for accessibility issues and fix them. The agent calls Agora’s scan_url tool and gets a score of 58 out of 100 with four issues: a button without a name, an image without alternative text, low-contrast text and an unlabelled input. It fixes the code in three files, scans again and reaches 100 out of 100.

Works with

  • Claude Code
  • Cursor
  • VS Code + Copilot
  • Windsurf
  • Claude Desktop
  • Codex CLI
  • Gemini CLI

The gap in AI-written UI

Agents write UI faster than anyone can review it for accessibility.

Coding agents learned from the web, and most of the web is inaccessible: icon-only buttons, missing labels, low-contrast text. Without a way to test, your agent cannot tell whether the component it just generated works for a keyboard or screen reader user. Now it can check, with the same engine your accessibility team already trusts.

Without Agora

  • The agent guesses at ARIA and hopes for the best.
  • Accessibility bugs surface in QA, in an audit or from your users.
  • Fixes happen weeks later, out of context and expensive.
  • “Looks fine to me” is the only verification.

With Agora MCP

  • The agent audits its own output in a real browser.
  • Every issue arrives with the selector, the HTML and how to fix it.
  • Fixes land in the same change, while the context is fresh.
  • A re-scan proves the fix before anyone reviews the code.

Set up in a minute

Connect once. Your agent does the rest.

  1. 01

    Turn it on

    The MCP server starts together with Agora and only listens on your computer. The AI Agents page shows its status and the address to use.

  2. 02

    Connect your agent

    Copy one command or config snippet from Agora, or use one-click install for Cursor and VS Code.

  3. 03

    Just ask

    “Scan localhost:3000 and fix what you find.” Your agent calls Agora, applies the fixes and verifies them.

Inside a tool call

What your agent actually receives

No vague scores. Agora returns every violation with its WCAG criterion, the exact element, its HTML and a concrete fix, in a format an agent can act on immediately.

Agent → Agora

{  "method": "tools/call",  "params": {    "name": "scan_url",    "arguments": {      "url": "localhost:3000/checkout",      "selector": "#payment-form",      "viewports": ["desktop", "mobile"]    }  }}

Agora → agent

58

http://localhost:3000/checkout

3 issues · 5 elements · WCAG 2.2 AA

  • Critical button-name WCAG 4.1.2

    Buttons must have discernible text

    button.pay-apple

    <button class="pay-apple"><svg …/></button>

    How to fix: Element does not have inner text that is visible to screen readers; aria-label attribute does not exist or is empty

  • Critical label WCAG 4.1.2

    Form elements must have labels

    #card-cvc

    <input id="card-cvc" placeholder="CVC">

    How to fix: Element does not have an explicit <label>; aria-label attribute does not exist or is empty

  • High color-contrast WCAG 1.4.3

    Elements must meet minimum color contrast ratio thresholds

    .total__hint

    <p class="total__hint">Taxes included</p>

    How to fix: Element has insufficient color contrast of 2.9 (foreground color: #a3a3a3, background color: #ffffff). Expected contrast ratio of 4.5:1

Four tools, every workflow

Test the page you are running, or the markup you just wrote

Scan any URL

scan_url

Local dev servers, staging, production or a file on disk.

  • Focus on one component with a CSS selector
  • Wait for client-rendered content
  • Desktop and mobile viewports
  • Optional screenshot of the scanned area

Scan raw HTML

scan_html

Audit a component before it is wired into a page, with no server needed.

  • Rendered in memory in a real Chrome
  • Component mode skips page-level rules
  • Loads your CSS from a base URL for accurate contrast
  • The same detailed report as scan_url

Validate attributes

validate_attributes

Scanners find missing names. This checks whether the names that are there actually make sense.

  • Checks aria-label, alt, title, labels and link text
  • AI flags vague values like “image” or “click here”
  • Catches file names and wrong-language labels
  • Selector, HTML and reason for each flagged value

Simulate a screen reader

simulate_screen_reader

Hear a page or component the way blind users do, straight from Chrome’s accessibility tree.

  • Every announcement in reading order
  • Headings, landmarks, links, forms, images and tables
  • AI review of reading order and heading structure
  • Works on a URL or on raw HTML

Accessible by default

Accessibility becomes part of how your agent works

Instead of a separate audit weeks later, testing happens inside the same loop your agent already runs for types, tests and linting.

  1. 01

    Build

    Your agent writes or changes UI.

  2. 02

    Scan

    It calls Agora on the page or component.

  3. 03

    Fix

    It applies the fixes Agora points to.

  4. 04

    Verify

    It scans again until the issues are gone.

Why teams choose it

Built for developers, trusted by accessibility teams

4
tools: two scanners, label checks and a screen reader
7+
AI assistants with guided setup
WCAG 2.2
A and AA rules, plus AAA checks
0
bytes of your code sent to the cloud

Frequently asked questions

What is MCP?
The Model Context Protocol is an open standard that lets AI assistants use external tools. Agora includes an MCP server, so any MCP-capable assistant can run accessibility scans through it.
Which AI assistants are supported?
Any client that supports MCP. Agora includes step-by-step setup for Claude Code, Cursor, VS Code with GitHub Copilot, Windsurf, Claude Desktop, Codex CLI and Gemini CLI.
Does my code leave my computer?
No. The MCP server only accepts connections from your own machine, and Agora renders and scans the pages locally. The AI checks in validate_attributes and simulate_screen_reader send only the collected labels and announcements, never the full page, to the QualiBooth API.
Can it test pages on localhost?
Yes, that is what it is built for. Point your agent at your dev server, a staging site or a local file.
Can it test a component that is not on a page yet?
Yes. The scan_html tool renders markup in memory, without a server. Component snippets skip page-level rules such as document title and landmarks, so the report focuses on the component itself.
What do I need to get started?
Agora, signed in with a QualiBooth account that has a desktop license. The MCP server starts automatically, and the AI Agents page walks you through connecting your assistant.
Does it replace manual accessibility testing?
No. Automated checks catch a large share of issues early, but keyboard, focus and screen reader behaviour still need human testing. Agora MCP makes sure the basics are right before a person looks.

Ship accessible UI at the speed of AI

Start a free trial of Agora, connect your coding agent in a minute and watch it fix accessibility issues before they reach code review.

Start your free trial Book a demo