Fetch one web page and return it as compact Markdown. Pages are fetched directly for free when the site allows it, rendered in a local headless browser (also free) when they need JavaScript, and fetched through Bright Data Web Unlocker (billed) only when the site blocks; a note after the result says which. Long pages are paginated: when truncated, call again with the returned offset. Use format=raw for the unprocessed Markdown.

Fetch up to 10 pages in parallel and return [{url, ok, source, content | error}] as JSON. Each page is capped (max_chars_per_url); fetch a single page with scrape_as_markdown to read past the cap. Pages are fetched directly for free when the site allows it, rendered in a local headless browser (also free) when they need JavaScript, and fetched through Bright Data Web Unlocker (billed) only when the site blocks; source says which.

Fetch one page and return its raw HTML, paginated with offset/max_chars. Prefer scrape_as_markdown unless you need markup, attributes or embedded JSON. Pages are fetched directly for free when the site allows it, rendered in a local headless browser (also free) when they need JavaScript, and fetched through Bright Data Web Unlocker (billed) only when the site blocks; a note after the result says which.

Fetch a page and turn it into JSON using your own model via MCP sampling (no extra Bright Data cost beyond the page fetch). Pass extraction_prompt and optionally a JSON Schema in schema to shape the output.

Usage this session: calls, errors, latency (p50/p95) and cache hits per tool, billable units, estimated cost, free direct fetches and savings, remaining budget, zone status and open browser sessions.