The #1 deep research agent

GPT Researcher plans, searches, reads and cites. Ask a question and get a detailed, source-backed report, from the web or your own documents, with any LLM. Fully open source under the Apache 2.0 license.

Get started GitHub ★ 30k

research.py

from gpt_researcher import GPTResearcher

researcher = GPTResearcher(
    query="Are solid-state batteries ready for EVs?",
    report_type="research_report",
)

await researcher.conduct_research()
report = await researcher.write_report()

Report ready · 18 sources · 57 citations

Built for real research

One agent for quick answers, long reports and everything in between.

Web research

Plans the questions, searches in parallel and reads 20+ sources per report, with a citation for every claim.

Your documents

Research over PDFs, Word, spreadsheets and Markdown, on their own or combined with the web.

Deep research

Explores a topic recursively, branching into sub-questions until the answer is thorough.

Multi-agent teams

LangGraph and AG2 agents that plan, research, review and publish together.

MCP sources

Bring GitHub, databases and your own APIs into research through the Model Context Protocol.

Smart context

Jev keeps only the passages that answer the question: 59% more relevant context than embeddings.

How it works

01
Plan
Breaks your question into focused sub-questions.
02
Search
Queries the web or your documents for each one, in parallel.
03
Filter
Keeps only the passages that answer each sub-question.
04
Write
Produces a cited report in Markdown, PDF or Word.

New

Jev’s context is 59% more relevant than embeddings, at the same cost

GPT Researcher now filters sources with Jev by default, and falls back to local keyword ranking, so no key or embeddings provider is required. Read the benchmark →

Cited in 150+ research papers

Researchers use GPT Researcher as a baseline, a building block and a subject of study.

All papers on Google Scholar ↗

Start researching in five minutes

Read the quickstart Join the Discord