Scraping fails in predictable stages: the plain fetch gets blocked, the proxy gets challenged, the challenge needs a browser, the browser meets a CAPTCHA. Most tooling makes you build that ladder. Here it is one call that climbs it, which means an agent does not have to reason about why a page came back empty. Note the timing on the Lighthouse tool — 30 to 120 seconds is long enough that a client with a short timeout will give up before it answers.
Browserless's endpoint. One tool is a stateful agent whose session persists across calls; the other eight are stateless REST-style operations. A separate Connector endpoint exposes a five-tool subset.
- browserless_agent — a multi-turn browser agent that handles logins and forms across calls
- browserless_smartscraper — cascading strategies: HTTP fetch, then proxy, then headless, then CAPTCHA solving
- browserless_function and browserless_download — run your own Puppeteer code, and capture what Chrome downloads
- browserless_export — HTML, PDF, image, or a full offline ZIP with resources
- browserless_search — web, news and image search via SearXNG, optionally scraping each result
- browserless_map and browserless_crawl — discover a site's URLs from sitemap and links, then scrape every page
- browserless_performance — a Lighthouse audit across accessibility, best practices, performance, PWA and SEO, taking 30 to 120 seconds
OAuth with a Browserless account, a bearer token, or a token query parameter — evaluated in that order, header first.
Paste a URL, then authorize — add the endpoint to your client, then approve the OAuth consent screen