deep-research
Run a deep agentic research task using anakin-cli that autonomously explores the web and returns a comprehensive report. Use for comprehensive research — comparisons, market analysis, technical deep-dives, or questions requiring multiple sources. Takes 1-5 minutes.
Deep research
Trigger
Comprehensive research on a topic — comparisons, market analysis, technical deep-dives, or questions requiring multiple sources.
Workflow
- Verify anakin-cli is authenticated by running
anakin status. - Create the output directory if it doesn't exist:
mkdir -p .anakin - Generate a descriptive filename from the topic:
- "React vs Vue comparison" ->
.anakin/research-react-vs-vue.json - "AI startup funding 2025" ->
.anakin/research-ai-funding-2025.json
- "React vs Vue comparison" ->
- Inform the user that deep research takes 1-5 minutes and runs autonomously.
- Run the research:
For complex topics that may take longer:anakin research "$ARGUMENTS" -o .anakin/research-<topic>.jsonanakin research "$ARGUMENTS" --timeout 900 -o .anakin/research-<topic>.json - Once complete, extract and present the key findings:
wc -l .anakin/research-<topic>.json && head -100 .anakin/research-<topic>.json jq -r '.generatedJson.summary' .anakin/research-<topic>.json - Present to the user:
- Executive summary
- Key findings and structured data
- Cited sources
- Offer to scrape any cited URLs for deeper analysis
Commands
# Deep agentic research (1-5 minutes)
anakin research "<topic>" -o .anakin/research-topic.json
# Extended timeout for complex topics
anakin research "<topic>" --timeout 900 -o .anakin/research-topic.json
Options
--timeout <seconds>— Max wait time (default: 600)-o, --output <path>— Save to file
Async behavior
Polls every 10s. Typical completion: 1-5 minutes.
Response
generatedJson with summary, structured_data, data_schema, and cited sources.
Guardrails
- Always warn the user about the 1-5 minute duration before starting.
- Use
/anakin:search-webfor quick factual lookups — reserve/anakin:deep-researchfor multi-source exploration. - Always save output to a file with
-oto prevent terminal buffer overflow. - If research fails, fall back to
/anakin:search-webplus multiple/anakin:scrape-websitecalls as a manual alternative. - If research times out, suggest increasing
--timeoutor breaking the topic into smaller queries.
Output
- Comprehensive research report with findings and sources
- File path where the full report was saved