GPT Researcher is a powerful, open-source AI agent designed for autonomous deep research. Using a multi-agent architecture, it breaks a complex question into subtopics, searches the web and/or your uploaded documents, summarizes key findings with citations, and produces a structured, exportable report.

It supports “deep research” mode for recursive exploration of subtopics, custom prompts for tone and structure, integration with vector databases, and multiple report output formats. Ideal for analysts, academics, journalists, founders, or anyone needing to generate longform, fact-rich content fast.

Inputs:

• A single research query or question

• Optional: Local documents (PDF, DOCX, MD, etc.) via DOC_PATH

• Optional: Custom prompt templates for tone, length, or format

• Optional: Vector store (e.g. Chroma, FAISS) for document recall

• Optional: API keys (OpenAI, Tavily, Serper, Browserless, etc.)

Outputs:

• Structured longform research report (2000+ words)

• Citations for all referenced sources

• Exportable formats: PDF, DOCX, Markdown, HTML

• Optional: intermediate results (.json, log.txt)

• Optional: charts/images pulled from web sources

Setup Instructions:

1. Install dependencies
   Requires Python 3.11+

   git clone https://github.com/assafelovic/gpt-researcher.git
   cd gpt-researcher
   pip install .

2. Set your API keys

   export OPENAI_API_KEY=your_key
   export TAVILY_API_KEY=your_key

3. (Optional) Add local docs folder

   export DOC_PATH="./docs"

4. Run the agent in code:

   from gpt_researcher import GPTResearcher
   
   async def run():
       researcher = GPTResearcher(
           query="How is climate change affecting Arctic wildlife?",
           report_type="deep"
       )
       await researcher.conduct_research()
       report = await researcher.write_report()
       print(report)

5. Or run from CLI:

   gpt-researcher --query "How is AI transforming education?" --report_type deep

   Pricing:

   • Free & open-source

   • Requires API usage from providers (OpenAI, Tavily, Serper, etc.)

   • No licensing cost or paid tier