A multi-agent research framework that produces cited reports from web and local documents
A planner agent generates research questions, executor agents gather and filter sources, and a publisher agent creates a cited report. Web and local document research, custom sources via MCP, recursive deep research, and export all fit within the same framework. An active project since 2023, it is one of the most mature general-purpose research frameworks.
General web research reports, comparisons of local documents and web sources, and building an embedded research engine.
Python 3.12, OpenAI/Tavily keys, and a large dependency set make setup challenging; API costs apply. “More sources = unbiased” is not true; independently verify web content against prompt injection and source claims.
npx skills add assafelovic/gpt-researcher