Crawl a website and audit every page for technical SEO issues: broken links, redirects, noindex, canonical, titles, meta descriptions, H1, hreflang, structured data and sitemap coverage. Schedule it to get only the pages that changed or broke since the last audit.
All notable changes to this Actor are documented here. The output fields are a contract: no field is
renamed, removed or retyped outside a major version.
[Unreleased]
Store page (2026-09-18)
Icon uploaded (the console asks to confirm the crop before the form can be saved), third category fixed to
AUTOMATION (MARKETING does not exist in the Store and was dropped silently).
README: what happens during a run, the fields returned for every page, an example run summary, and two
questions on staging sites and on run duration and compute cost.
Changed (2026-09-17, build 0.1.5)
maxPagesPerSite is capped at 2,000 (was 10,000): 1,000 pages were measured at 618 MB on the platform, and
larger audits were not verified within the default 1 GB.
Fixed (2026-09-17, build 0.1.4, tag beta)
A crawl cut short by the maximum cost per run now says so (run message, site detail, budgetReached in
SUMMARY) instead of reporting that the site had more pages than the page limit. Verified on the platform
with a $0.05 limit: 10 pages saved and charged, then a clean stop.
Fixed (2026-09-17, build 0.1.3, tag beta, private)
Dual-stack sites could fail on the platform with an empty ETIMEDOUT (robotstxt.org): Node gave each
IPv6/IPv4 connection attempt 250 ms, and the platform has no IPv6 route. Attempts now get 2.5 s, for page
requests, robots.txt and sitemaps alike.
A network error with an empty message is now reported with its code (fetchError, run summary).
Fixed (2026-09-17, build 0.1.2, tag beta, private)
Memory: build 0.1.1 ran out of 1 GB after 240 pages. With the Cheerio crawler, the crawling context held
each page's DOM, and the timers the request queue starts during a request kept that context alive under
Node 24. The crawl now uses a plain BasicCrawler that streams each response: the HTML lives only in the
handler, non-HTML bodies are not downloaded, HTML is capped at 5 MB, and each page is parsed once instead
of twice. Measured locally: heap flat at about 57 MB over 120 pages, instead of about 1.5 MB more per page.
Added (2026-09-17, version 0.1, not published yet)
HTTP crawl of one or more websites from a start URL, without following redirects, within the site's host
(www. and the bare domain are one site; subdomains on request), robots.txt always respected, identified
as TechnicalSeoAuditBot.
Sitemap discovery from robots.txt, or /sitemap.xml; sitemap-only pages are audited after the link crawl.
One item per URL: status, redirect, response time, indexability and reason, title, meta description, H1,
canonical, robots meta and X-Robots-Tag, lang, hreflang, JSON-LD types, word count, link counts,
broken internal links, inlinks, sitemap presence, issues with severity, score.
37 issue codes, including site-wide checks: duplicate titles and descriptions, orphan pages, sitemap
entries that are not indexable, pages missing from the sitemap.
Change tracking between runs (trackChanges, onlyChanges): new, changed, unchanged and removed pages,
with changed fields, previous values, new and fixed issues. Removed pages are only reported after a
complete crawl.
SUMMARY record per run: per-site status (ok, unreachable, disallowed, redirected_away,
not_reached), average score, top issues. Blocked or disallowed sites are not charged and do not fail the
run; the run fails only if most HTML pages come back with no title and no link (parser check).
Pay-per-event ready: the cost limit is checked before every saved item.
