Snapshot
Crawls a public site, screenshots every page, and exports the run as labelled evidence.
Use it when you need the before picture of a site ahead of a redesign, a migration or a client onboarding.
Access
This tool runs behind a sign-in. If you have access, open it; if not, ask and we will say yes or explain why not.
What it does
Maps a public site, captures each page at desktop, tablet and mobile widths in one run, builds a copyable architecture tree, and exports the whole run as labelled evidence. It is the before picture for a redesign, a migration, or a client onboarding, and the visual record that a later run can be compared against.
One codebase, two capture engines: local Playwright for day-to-day use and Cloudflare Browser Rendering for the hosted deployment. The crawl result records which engine produced it.
Features
- Crawl mode, an explicit page list, or a knowledge-base mode
- Desktop, tablet and mobile in one run, each viewport saved before the next starts so the browser holds one at a time
- Full-page or viewport screenshots
- capture-run, a terminal driver over the app's own endpoints, so a CLI run and a form run write identical output; it checks every requested page came back
- Directory tree view and copyable architecture tree text
- External dependency detection, from real network requests locally and from rendered HTML on Cloudflare
- JSON, CSV, full-run ZIP and labelled screenshot ZIP exports, generated statelessly from the current crawl result
- The whole tool sits behind one password, held as a signed session cookie; the API answers 401 and /crawls redirects without it
- Runtime chosen in the form. Auto prefers CAPTURE_PROVIDER, else Cloudflare when its credentials are present, else local Playwright; an explicit choice is honoured or refused, never swapped. The crawl result records which engine ran
Inputs
- Root URL. Crawled to a page and depth limit, or paired with a list of specific pages
- Crawl limits. Up to 100 pages and 10 levels deep running locally; a hosted crawl is capped at 6 pages and 180 seconds, and says so in the result
- Viewports. Any of desktop 1440, tablet 768 and mobile 390, captured one after another
- Capture options. Full page or viewport, query parameters, include and exclude patterns
- Help Center URL. Knowledge-base mode: a Zendesk Help Center's articles instead of screenshots
Outputs
- Screenshot gallery. One capture per page, with status and depth; files named from URL and viewport
- Saved run. Locally, captures/<date>-<host>/ with a manifest and an index.html contact sheet. Hosted, each viewport downloads as a ZIP instead
- Page tree and site architecture. Copyable as text
- Third-party dependency map. External domains each page loads from
- Crawl result (JSON)
- Page list (CSV)
- Labelled screenshots (ZIP)
- Full run (ZIP)