User Agent Parser / Analyzer

Know your audience’s hardware before you blame it. User Agent Parser reads the user-agent string out of access-log lines and classifies each request by browser, operating system and device type — with bots flagged separately so they don’t skew your picture of real human traffic.

Finding the user agent itself is the first step, and it works two ways. In standard combined-log format the UA sits in the last quoted field of the line (after the request and referrer fields), so the parser takes the final quoted segment and checks it looks like a real UA — containing telltale tokens like "Mozilla", "curl", "bot", "spider" or "http" — before trusting it. If a line doesn’t have that structure at all (for example you pasted bare user-agent strings, one per line, with no surrounding log fields), the parser falls back to treating the whole line as the UA. Either way, a line that produces nothing usable is counted as unparsed rather than silently dropped, so the "unclassified" number in the report tells you how much of your paste didn’t fit either shape.

Classification runs bot detection first: a line matching known crawler and tooling signatures (googlebot, bingbot, Slurp, DuckDuckBot, Baiduspider, Yandex, social-media link previewers, curl, wget, common HTTP client libraries) is flagged as a bot immediately and skipped for browser/OS classification, because "Chrome" inside a spoofed bot string would otherwise pollute your human browser stats. Everything else goes through simple, readable keyword rules for browser (Edge, Opera, Firefox, Chrome, Safari, in that priority order since some UAs contain multiple tokens) and OS (iOS, Android, Windows, macOS, Linux), plus a mobile/desktop split based on device keywords in the string.

• Browser classification — Chrome, Firefox, Safari, Edge and Opera, detected with priority rules that handle UAs containing multiple overlapping tokens.
• OS classification — Android, iOS, Windows, macOS and Linux.
• Device type — mobile vs. desktop split, plus a dedicated bot category.
• Bot detection — googlebot, bingbot, common crawlers and HTTP client libraries flagged and excluded from browser/OS stats.
• Top raw user agents — the exact UA strings seen most often, with counts, for evidence beyond the categories.

1. Paste access log lines (standard combined-log format) or bare user-agent strings, one per line.
2. Click Parse. Browser, OS, device and bot categories populate along with counts.
3. Skim the device mix and bot volume in the stat tiles for a quick read before diving into the full report.
4. Copy the report to attach to a front-end compatibility ticket or traffic-quality review.

Example 1 — Mobile vs. desktop split. If the report shows 70% mobile traffic, that changes what the front-end team should prioritize testing before the next release — a decision that’s hard to make from raw log lines but obvious from the classified counts.

Example 2 — Bot audit. Classify a day of traffic to see how much of your "traffic" is actually googlebot, other search crawlers, or scrapers rather than real visitors — important context before drawing conclusions from any raw request-count metric.

Example 3 — Confirming a device-specific bug. A bug report says "crashes on iPhone"; running the affected user’s requests through the parser confirms the OS and browser class (iOS Safari, say) match the report, ruling out a red herring.

Example 4 — Spotting spoofed or unusual clients. The "top raw user agents" list surfaces the exact strings behind your traffic — an unfamiliar or malformed UA appearing frequently near the top is worth investigating even if it doesn’t match a known bot signature.

• Audience clarity — see your real browser, OS and device mix at a glance.
• Bot visibility — crawlers and scrapers are separated out, not mixed into human traffic stats.
• Exact evidence preserved — top raw UA strings back up the categorized counts.
• Works on raw logs or bare UAs — no need to pre-extract the field yourself.
• Private — parsing and classification happen entirely in your browser.