Gpt Researcher

GPT Researcher is an open source autonomous research agent. You give it a question, and it plans the research, reads many sources in parallel and writes a detailed report in which every claim traces back to a citation. It researches the public web and your own local documents, and it outputs markdown, PDF, docx or a plain string. It is built for developers and teams who want deep research inside their own stack rather than through a hosted product. The architecture is a planner that splits a question into research questions, one researcher per question running at the same time, a context filter that keeps only the passages answering the question before anything reaches the model, and a writer that assembles the report. More than 20 retrievers are supported, including Tavily, Google, Bing, Exa, arXiv and PubMed, alongside any MCP server. Any model provider can be used, from OpenAI and Anthropic to Groq, Llama 3 and models hosted locally. A deep research mode lets findings seed new questions so a run explores a tree of subtopics instead of a single pass. A companion Model Context Protocol server exposes deep research, quick search, report writing and source retrieval as tools to Claude, Cursor, n8n and any other MCP client. Installation is a single pip install. Pricing is $0. The project is published under the MIT license with no subscription, no API gateway and no rate limit imposed by the project itself. You host the agent and pay your chosen model and search providers directly. The vendor gives an indicative range of a few cents to a few dollars of provider spend for one deep research run on default settings. There is no hosted service today, and managed deployments inside a customer VPC are arranged by direct contact.