research

Autonomous deep research — multi-step investigation that browses, verifies, and compiles a cited report. These calls are slower than web (seconds to tens of minutes) and stream progress to stderr unless quieted.

seek research <provider> "<query...>" [options]

Providers: brave, parallel, google, perplexity.

research brave

Brave Answers Deep Research — a quick, surface-level pass. Uses BRAVE_ANSWERS_API_KEY.

seek research brave "what changed in the EU AI Act this month?" --citations

Option	Description
--country <cc>	Two-letter country code (default us)
--language <code>	Response language (default en)
--citations	Request provider citation tags
--entities	Request provider entity tags
--json	Emit raw JSON response

Output: report text with a citations list appended.

research parallel

Parallel Deep Research — thorough, from ~30s up to ~25 minutes depending on processor depth. Supports strict JSON output via a schema file, giving auditable sources for B2B and data-ops work.

seek research parallel "list the top 10 EV battery manufacturers" \
  --schema ./schema.json --processor pro

Option	Description
--schema <path>	Path to a JSON file describing the desired output schema
--processor <name>	Depth: core-fast | core | core2x-fast | core2x | pro-fast | pro | ultra-fast | ultra (default core)
--json	Emit raw JSON response

Processor guidance: -fast variants trade data freshness for speed. core = cross-referenced moderate complexity; core2x = high-complexity cross-referenced; pro = exploratory web research; ultra = advanced multi-source deep research.

Output: report content (JSON-stringified if a schema was used) with unique citation URLs appended.

research google

Gemini Deep Research Agent — comprehensive, 5–20 minutes. It autonomously plans, executes, and synthesizes multi-step research into a detailed, cited report. Uses GEMINI_API_KEY.

seek research google "summarize the SEC investigation into ACME Corp"
seek research google "map the competitive landscape for open-weight speech models" \
  --interactive --planner perplexity

Option	Description
--agent <name>	Override the Deep Research agent id
--system <text>	Optional system instruction (e.g. “act as an equity analyst”)
--interactive	Review and approve a research plan before starting
--planner <provider>	Planner for --interactive: perplexity | brave | manual (default perplexity)
--quiet	Suppress polling progress on stderr
--json	Emit raw JSON interaction object

--interactive requires a TTY; it proposes a plan you can approve, revise, or quit before the long-running research begins.

Output: report text with a citations list appended.

research perplexity

Perplexity Sonar Deep Research — an informed pass, ~1–3 minutes, run as an async job. Uses PERPLEXITY_API_KEY.

seek research perplexity "state of the small language model market"

Option	Description
--model <name>	Override model (default sonar-deep-research)
--effort <level>	Reasoning effort: low | medium | high (default medium)
--quiet	Suppress polling progress on stderr
--json	Emit raw JSON async job object

Output: report text with a citations list appended.