Paste competitor product URLs, the Agent crawls them every morning via `vb crawl` and lands a clean per-SKU price CSV in your Drive. The monthly trend PDF flags meaningful moves automatically — 'competitor A cut Pro by 12% on May 3, undercutting our list by $30'.
Transient price experiments are excluded from the smoothed line by default. The annotated pins mark only moves that held for at least 3 consecutive crawls — no noise-triggered panic alerts.
When procurement or legal says "the vendor disputes the price six months later", the Drive archive proves what was on the page that day.
Daily 07:30 · 23 URLs · daily CSV + Sunday digest
One URL per line. Don't worry about price selectors — the Agent reads the page and picks the price out of the rendered HTML / JSON structured data on its own. If parsing fails (page redesign, geo-block), the Agent surfaces the failure in chat instead of writing a wrong number. Product pages, listing pages, region-specific storefronts all work.
Tell the Agent in chat — "crawl these every morning at 7:30, weekly digest Sunday night, monthly trend PDF on the last day" — and the Schedule is set. The URL list is written to `watchlist.yaml` in Drive; adding a new SKU is one chat message that edits the file.
Every month's trend PDF lands in /price-monitor/monthly/ in Drive. Per-SKU charts with notable moves annotated, a competitive landscape table, and a Schedule card showing exactly when the data was collected. The raw CSVs sit underneath so your analyst can pivot however they want.
Cheap price-monitor tools fall back to a stale regex when a site redesigns and silently capture wrong values. The Agent reads the rendered page each day; when it cannot find a price with confidence (page redesign, geo-block, missing element), it surfaces the failure in chat and leaves that day's row empty rather than fabricating. You catch it the same morning, not three months later when procurement disputes a number.
Set the daily Schedule once and the Agent runs in the background. Output lands in Drive. No interactive checking, no manual triggering, no "did the crawl run today". You only look at the file when the weekly digest or monthly PDF flags something material.
Every captured price links to a snapshot of the source page (rendered HTML + screenshot) at the moment of collection. If a vendor later disputes a price you screenshotted three months ago, the Drive archive proves it. The procurement / legal team can rely on the data.
The URL list and parsing rules live in a plain `watchlist.yaml` in your Drive. You can edit it in chat ("add this new SKU", "stop tracking the discontinued one", "shift the crawl to 6am") and the Agent rewrites the file; the next run reads it back. No locked-in subscription — the config is yours, plain text, owned in your Drive.
`vb crawl` supports both fast HTML fetch and headless-browser rendering (for JS-heavy pages). It respects robots.txt and rate-limits politely. We don't fingerprint-evade hostile sites — if a competitor explicitly blocks scrapers, the Agent will say so in chat and ask you to confirm whether to back off. We do not run abusive scraping campaigns to bypass intentional blocks; the law and platform terms decide what's allowed.
Sign in, hand it over to the Agent — the finished file lands in your Drive.