GPT Researcher

Give GPT Researcher a query and it plans research questions with a planner agent, dispatches execution agents to gather sources, then aggregates everything through a publisher into a cited report, typically exceeding 2,000 words and drawing on more than 20 sources. The original open deep research agent, it also searches local documents including PDF, Word, Excel, CSV, Markdown, and PowerPoint files, and a deep research mode explores topics in a tree-like pattern at an estimated cost of about 40 cents per run. MCP integration connects specialized data sources, Google Gemini can generate inline report images, and two frontends ship in the repo: a lightweight static one and a production NextJS app. Installation is pip install gpt-researcher plus your model and search API keys, or Docker for the full stack, and the package embeds cleanly into other applications, which is why many later research stacks build on it. Licensed Apache-2.0 with around 28.7k stars and more than 3,000 commits, it is maintained by Assaf Elovic with documentation at docs.gptr.dev.