Point Dr Urls at a URL and it returns a ranked, fix-by-fix audit — SEO, performance, security and accessibility — as Markdown your agent applies directly. It finds what's wrong; Claude Code, Cursor or your own agent writes the fix.
Left is the problem in your words. Right is what the scan does about it.
“Your coding agent can change code but has no idea what's actually wrong with the live site”
POST a URL and get back exactly what's wrong, ranked by impact — the agent starts from findings, not guesses.
“You paste screenshots and vague complaints into the agent and hope it guesses right”
The Markdown brief names each problem, why it matters, and the exact change, so there's nothing to describe by hand.
“Lighthouse gives a score, not a task list an agent can act on”
Output is a prioritised list of concrete fixes with code examples, not a single number.
“Every audit tool outputs a dashboard for a human, not structured data for a model”
format:"markdown" returns an agent-ready document; the default JSON is there when you'd rather parse fields.
“The agent fixes one thing and silently breaks indexing or a security header”
Re-audit after the agent's change and diff the two runs — a regression in any category shows up immediately.
“You want the audit in your pipeline, not in yet another browser tab”
It's one authenticated HTTP call, so it drops straight into a script, a CI step, or an agent tool call.
POST a URL to /api/v1/check with format:"markdown" (or Accept: text/markdown) and you get a ranked, fix-by-fix brief back as text/markdown. JSON is the default if you'd rather read individual fields. Try it on the free check page and click Copy for AI to see the exact document.
Not yet — that's on the public roadmap. Today the integration is the HTTP API plus the Markdown brief, which already works with any agent that can make a request or accept pasted context. A native MCP server that exposes the audit as first-class tools is the next step.
Because most of what it finds IS code: missing meta tags, broken canonicals, unlabelled controls, render-blocking scripts, absent security headers. These are things an agent can fix in the source — it just needs to be told precisely which ones, which is what the brief provides.
One page is free and instant, no account. Open one and I'll take the whole site and keep every check-up on record.