User-agent parser

Six fields, printed as JSON: browser, browserVersion, os, engine, deviceType and isBot. Anything it does not recognize comes back as null rather than as a guess, so a null means the pattern did not match, not that the string lacked the information.

By pattern, in a fixed order that matters. Edg is checked first, then OPR, Firefox, CriOS, Chrome, then Safari, then Internet Explorer. That order exists because every Chromium browser carries Chrome in its string and Safari carries it too. Edge is reported as Edge, not Chrome, precisely because Edg is tested first.

From two patterns and a fallback. Anything matching Mobi, or Android together with Mobile, is mobile. Anything matching Tablet or iPad is tablet. Everything else is called desktop, including strings it did not recognize at all. Recent iPads that present themselves as desktop Safari are therefore labeled desktop, which is what their string says.

It is regex and table based, with no maintained device database behind it. Vendors change their strings often, so a new or obscure agent can come back with nulls instead of matches. It errs toward reporting nothing rather than naming the wrong browser. For analytics at scale, a maintained database is the right tool.

One case-insensitive pattern across the whole string, looking for bot, crawl, spider, slurp or bingpreview. That catches Googlebot, bingbot and most crawlers that identify themselves, and it misses anything disguised as a normal browser. isBot true is a useful signal; isBot false proves nothing.

No. It stays in this tab and is matched against the patterns inside a Web Worker. That matters because a UA string pulled from a log can identify a specific device build, and this tool never transmits it anywhere or keeps it after the run.