Web Search and Crawl

Choose the right operation

Task	Command	Credit cost
Find pages and inspect results	anycap search --query	1
Get a synthesized cited answer	anycap search --prompt	5
Read one known URL as Markdown	anycap crawl <url>	1

Structured search

anycap search --query "Go error handling" --no-crawl --max-results 10 \
  | jq -r '.data.results[] | "\(.title) — \(.url)"'

Remove --no-crawl when you need full Markdown content for every result. Filter by freshness, exact dates, or domains:

anycap search --query "agent releases" --time-range month
anycap search --query "language update" --after 2026-01-01 --before 2026-06-30
anycap search --query "context patterns" --include github.com --exclude pinterest.com

Do not combine --time-range with --after or --before; exact dates take precedence.

Grounded search

anycap search --prompt "What changed in the latest Go release?" \
  | jq -r '.data.content'

Sources and per-segment citations are available under data.search_metadata when the grounding model decides to search.

Crawl a page

anycap crawl https://go.dev/doc/ \
  | jq -r '.data.markdown'

Crawl extracts primary HTML content as Markdown. It does not download binary files and may not capture content rendered only after client-side JavaScript runs.