Docket / About

About Docket

Docket is built by Matt Kerr, one person, in the UK. It is a Mac app that audits a site's SEO, copy and branding against 93 checks across 13 lanes, runs the crawl on your own machine, and costs $149 once — v0.1.0 is free while it is in beta.

That is the whole company. There is no team page because there is no team, no office address because there is no office, and no support desk because there is one person reading the issue tracker. Saying so is more useful than a stock photo.

Why it exists

Every SEO tool worth using is a subscription to somebody's crawler farm. You paste in a domain, a queue somewhere fetches it, and you pay monthly for the privilege of being told what your own HTML says. The crawl is the cheap part. The recurring bill is for the servers, and the servers exist because the tool was built as a service rather than as software.

Docket is the other shape. The audit engine is Python with no third-party dependencies at all, frozen into the app, and it runs on your Mac. There is no account, no crawl credit, no per-seat price and no queue. Audit the same site forty times in an afternoon; nobody is counting, because there is nobody to count.

What that changes about the findings

Running locally is not only a pricing decision. A crawler farm sees your site as an anonymous datacentre IP; your Mac sees it the way a person does, and can be told to introduce itself as any crawler you like. That is how Docket can tell you what your CDN returns to GPTBot specifically, rather than what your robots.txt claims it should.

It is also why the awkward findings are possible. Docket will tell you your AI opt-out is unenforceable, that your directive names a crawler that has been retired, or that the markup Google needs sits past the 2MB it will actually read. We publish those measurements as datasets rather than as marketing: 6236 sites' robots.txt files in the directives survey, 85 homepages in the entity survey, 82 in the page-weight survey. The raw JSON for each is in the site's repository, next to the scripts that check every figure on this site is derived from it rather than typed into a sentence.

What Docket is not good at

The honest list, because you will find it out in the first hour anyway:

How you can check any of this

The download is signed with an Apple Developer ID and notarised, which means the certificate carries a legal name and Apple has checked it. You do not have to take that on faith. After downloading the 17 MB disk image:

codesign -dv --verbose=4 /Volumes/Docket/Docket.app
spctl -a -t open --context context:primary-signature -v Docket-0.1.0-arm64.dmg

The first prints the authority chain, ending at Apple's root. The second should say accepted and source=Notarized Developer ID. Any Mac app worth installing will survive both; a surprising number will not.

What to do if a finding is wrong

Tell us, with the URL. Docket has shipped false positives — the byte-cap check once reported a major news site's title tag as past Googlebot's cutoff, when what it had found was an inline SVG icon label reading "Close icon" and the real title was fine. That is written up in full on the page about the check, because a tool that hides its mistakes is asking you to trust its output more than its authors do.

How to get in touch

Common questions

Who makes Docket?

Matt Kerr, one person, in the UK. There is no team and no company office. The macOS download is signed with an Apple Developer ID and notarised, so the legal name on the certificate can be checked with codesign before you install anything.

Does Docket send my site data anywhere?

No. The crawl runs on your Mac and results are stored in ~/.docket/ as plain JSON. Two optional connectors do reach outside: the knowledge refresh fetches one public file from docketseo.app and tells it nothing about what you are auditing, and PageSpeed Insights — off unless you add your own Google API key — necessarily sends Google the URL you asked it to measure.

What can Docket not do?

It is macOS only on Apple Silicon, it does not track keyword rankings or backlinks, and it is not built for crawling hundreds of thousands of URLs. The application source is not public, though the website's datasets are.