Find, fix, and verify accessibility issues with your AI coding agent
Accessibility work runs on a slow feedback loop. You read an audit report, fix the issues in your code, deploy, and wait for the next audit to learn whether any of it worked. The A11y Pulse MCP server hands that whole loop to your AI coding agent, from finding the issues across your site to proving they’re fixed.
The problem: the audit and the work live in different places
The reason the loop is so slow is that every part of it lives somewhere different. The issues sit in an audit tool, the fixes happen in your codebase, and the confirmation only exists in the next scan’s results, if you remember to go back and check.
That fragmentation is why accessibility backlogs grow: every issue carries the overhead of hopping between tools, and verification gets deferred until somebody remembers to look. Your AI coding agent could take on most of this work, but until now it had no access to the audit data and no way to confirm its own fixes.
The solution: your agent runs the whole workflow
We built the A11y Pulse MCP server to close that gap. It’s a hosted connector that gives your AI coding agent access to your A11y Pulse account: the full scan results for each of your sites, prioritised by impact and paired with the domain knowledge to trace each flagged element back to the right source file.
And as of this week, your agent can act on that account too. It can add new sites and pages, update site settings, trigger a scan, and wait for the results. Put those together and you get a complete workflow:
Add /checkout/confirm to A11y Pulse and scan it. Fix anything critical, then scan again to confirm.
Your agent registers the page, queues a scan, walks through the violations it finds, and proposes fixes grouped by source file for your review. Nothing is applied without your approval, and nothing is committed by the agent. Once you’ve shipped the fixes, it runs a fresh scan and reports back with the result, so the session ends with confirmation that the issues are resolved rather than a reminder to check a dashboard later.
This is also the difference between A11y Pulse and pointing a one-off audit tool at a URL. Single-page checkers tell you what’s wrong with the page in front of you. Your agent working from stored, site-wide audit data can tell you what matters most across the whole site, fix it, and prove the fix landed.
You decide what your agent can do
Giving an agent the ability to change things in your account is a bigger ask than letting it read, so that part is entirely up to you. Your API key carries the permissions you choose: Read access lets an agent fetch scans and suggest fixes, Manage sites & pages lets it add and update sites and pages, and Trigger scans lets it start scans on demand. Grant all three for the full workflow, or hand your agent a read-only key if you’d rather it only fetched and fixed. Deleting sites and pages isn’t possible through the MCP server at all; that stays in the dashboard and API on purpose.
Getting started
The server is hosted at mcp.a11ypulse.com, so there’s nothing to install. Create an API key, then copy the config snippet for your tool from Settings → Integrations → MCP Server. It works with Claude Code, Cursor, and GitHub Copilot, and setup takes a couple of minutes. The MCP server docs cover every tool and some prompts to start with.
Try it now
Start a free trial at a11ypulse.com, add your site, and connect your agent. By the end of your first session it will have found your accessibility issues, fixed the ones you approve, and shown you they’re gone.
Not using A11y Pulse yet? Sign up for a free trial and see how easy it is to bring continuous accessibility testing into your team's workflow.
Questions? We would love to hear from you. Drop us a line at [email protected].