What it does
An autonomous research agent that decomposes any research query into sub-questions, gathers information from web sources and local documents, and produces detailed factual reports with citations. It orchestrates a multi-agent workflow: a planner generates targeted research questions, crawler agents retrieve and summarize relevant information from 20+ sources, and a publisher aggregates findings into a comprehensive report. Supports image scraping, multiple output formats (PDF, Word), and plugs into any LLM provider via configurable backends.
Who it's for
Researchers, analysts, and investigators who need comprehensive, factually-grounded reports without weeks of manual gathering. Product managers evaluating market conditions, journalists fact-checking claims, business development teams conducting due diligence, or anyone needing to aggregate information from many sources quickly and with attribution.
Common use cases
- Generate detailed market research and competitive analysis reports
- Conduct due diligence research on companies, technologies, or investments
- Fact-check claims and gather supporting evidence from multiple sources
- Background research for investigations or journalism with source attribution
- Aggregate findings from 20+ web sources into a single authoritative report
Setup pitfalls
- Requires Python 3.11 or later; will fail silently on older versions.
- Three secrets detected in the codebase — store API keys (OpenAI, Tavily, and LLM provider credentials) in
.env, never hardcode or commit them. - High risk class due to filesystem writes and network calls — verify output directory permissions and review what it downloads; hostile research queries could attempt to write or exfiltrate data.
- Tavily API key required for web search; local-only research is possible without it but limited to provided documents.