Search options

Control providers, ranking, retrieval, concurrency, and output.

Search options

Choose a search engine

Kestrel Search supports DuckDuckGo, Bing, and Yahoo.

kestrelsearch search "python typing" --engine bing

Use fallback mode

Try providers in order until one succeeds:

kestrelsearch search "python typing" \
  --engine duckduckgo \
  --engine bing \
  --mode fallback

Search providers concurrently

kestrelsearch search "python typing" \
  --engine bing \
  --engine yahoo \
  --mode fanout

Run multiple queries

Repeat -q or --query. A positional query is optional when at least one query option is present:

kestrelsearch search \
  -q "python typing" \
  -q "pyright documentation" \
  --mode fanout

You can also combine the forms, for example kestrelsearch search "python typing" -q "pyright documentation".

Return recent results

Use d, w, m, or y for day, week, month, or year:

kestrelsearch search "Python news" --time-filter w

Skip page retrieval

For a faster snippet-only search:

kestrelsearch search "OpenAI news" --no-fetch

Return JSON

kestrelsearch search "Rust ownership" --output json

Control result count

kestrelsearch search "climate change" --top-k 3

Tune retrieval resources

Search requests, page downloads, and HTML parsing have separate concurrency limits:

kestrelsearch search "machine learning" \
  --search-concurrency 4 \
  --concurrency 8 \
  --parse-concurrency 4

Bound the number and size of fetched pages:

kestrelsearch search "machine learning" \
  --top-k 5 \
  --fetch-candidates 10 \
  --max-response-bytes 1000000 \
  --content-limit 3000

--fetch-candidates defaults to three times --top-k. Increasing it may improve ranking recall but also increases network, parsing, and downstream token costs.

Run the following command for the complete option reference:

kestrelsearch search --help

See the CLI reference for every option and default.