In one line: a free tool that turns a Google PageSpeed Insights / Lighthouse audit into a structured CLAUDE.md fix brief that the Claude Code AI agent can act on directly. No account, no signup, nothing leaves your machine except the URL you test.
What it does
You give it a URL. It runs Google PageSpeed Insights, then generates a CLAUDE.md brief containing every failing Lighthouse audit with prescriptive fixes, ranked offenders, the throttling profile used, third-party impact, a main-thread breakdown, and a "do not regress" list of currently passing audits. You drop that file into your repository and let Claude Code do the work.
Why it exists
The product itself is small on purpose. The thesis behind it is not. Most engineering tools (Lighthouse today; error monitors, type checkers, and security scanners next) are excellent at telling you what is wrong and almost silent on what you will break if you are careless. For an AI agent, that second half is the whole job. This is an AI context layer: it turns structured machine output into agent-ready briefs, with the guardrails written down as plainly as the task.
Where you can use it
- Web app at lighthouse-md.com, the primary surface.
- Chrome Extension, PageSpeed → CLAUDE.md on the Chrome Web Store, audits the active tab in one click. Built because a user asked for it.
- Audit reference library at /audits/, plain-English explainers for the Lighthouse audits that fail most often, mirrored as raw markdown for AI and LLM crawlers.
Who builds it
Made by Baris Kocdur. It started as a weekend project to stop copy-pasting PageSpeed audits into Claude, and it has grown from user feedback in public, one suggestion at a time. It is free and will stay free; there is no account and no data collection beyond the URL you submit (see the privacy policy).
Open to feedback
If you do SEO or web-performance work with Claude and something is missing, the fastest way to change the roadmap is to say so. Reach the maker on LinkedIn.