scrape
Targeted page extraction — clean text, Markdown, or JSON from URLs you already have. Pick a provider subcommand.
seek scrape <provider> <url...> [options]Providers: exa, parallel, firecrawl.
scrape exa
Bulk parsing — retrieves clean, stripped page text optimized for LLM token windows. Accepts multiple URLs.
seek scrape exa https://example.com https://example.org| Option | Description |
|---|---|
-c, --chars <n> | Max characters per page (default 50000) |
--json | Emit raw JSON response |
Output: one Markdown block per URL (title, URL, text), separated by ---.
scrape parallel
Enterprise extraction via Parallel’s /v1/extract endpoint — clean markdown
excerpts (and optional full content) plus publish dates per URL. Can
authenticate through logins and paywalls. Accepts multiple URLs.
seek scrape parallel https://app.example.com/dashboard \
--objective "Extract the latest invoice table"| Option | Description |
|---|---|
--objective <text> | Focus excerpts on a natural-language extraction goal |
--full | Include full markdown content, not just excerpts |
-c, --chars <n> | Max characters per excerpt block |
--max-total <n> | Max characters total across all URLs |
--max-age <secs> | Reuse cached content up to N seconds old |
--json | Emit raw JSON response |
Output: one block per URL (title, URL, publish date, content), with any per-URL errors listed at the end.
scrape firecrawl
Turn a single live URL into clean Markdown, HTML, JSON, a summary, or a screenshot via Firecrawl v2.
seek scrape firecrawl https://example.com -f markdown,links| Option | Description |
|---|---|
-f, --format <list> | Comma-separated: markdown, html, rawHtml, links, summary, screenshot, video, audio (default markdown) |
--json-prompt <text> | Add a JSON format with this LLM extraction prompt (produces data.json) |
--full | Include nav/footer/sidebars (disables onlyMainContent) |
--wait <ms> | Wait N ms for JS to render before scraping |
--mobile | Render as a mobile viewport |
--block-ads | Block ads / trackers during render |
--proxy <mode> | Proxy mode: basic | stealth | enhanced | auto |
--max-age <ms> | Reuse cached scrape up to N ms old (cheaper, near-instant) |
--json | Emit raw JSON response |
Takes a single URL. Output: the first available rendered format (markdown, then summary, html, links, JSON, or screenshot).