What is GPT Researcher?
GPT Researcher, started by Assaf Elovic in May 2023, is an open-source research agent. A planner breaks the question into research questions and execution agents search, scrape and summarise sources for each. A publisher then writes a cited report, often over 2,000 words, exported to Markdown, PDF or Word. It runs as a FastAPI server with a NextJS front end, as a Docker Compose stack, or as the gpt-researcher Python package. Local documents (PDF, Word, Excel, PowerPoint, CSV, Markdown, text) can replace or add to web sources, and MCP servers can be added as retrievers.
The repo has about 30,400 GitHub stars. GitHub lists it as Apache-2.0, though the PyPI package metadata says MIT. v3.7.0 was released on 26 September 2026 alongside PyPI 0.16.1, which had about 6,700 downloads in the latest seven days. The default stack needs an OpenAI key and a Tavily search key; any OpenAI-compatible endpoint can stand in for OpenAI. The project's MCP server now lives separately in assafelovic/gptr-mcp.
Is GPT Researcher worth it?
GPT Researcher is the most established open alternative to the deep-research modes in ChatGPT, Gemini and Perplexity, and the one most often embedded in other products. Its value is the pipeline rather than the UI: question planning, parallel scraping, per-source summaries with source tracking, and a long report with citations. All of that is callable in three lines of Python. A recursive Deep Research mode explores subtopics in a tree. The README puts it at about five minutes and about $0.40 per run with o3-mini on high reasoning effort.
The costs are low, but the numbers are the project's own. The README's context-filter benchmark, run by the maintainers on 28 research tasks, puts a standard report at about $0.115 with filtering and $0.192 without it. The default filter is Jev, a hosted scoring model from TypeSafe that needs a TYPESAFE_API_KEY. Without one, it falls back to local BM25 keyword ranking, so nothing extra is required, but the headline relevance figure (73% of kept passages relevant against 46% for embeddings) applies only to the paid option.
Out of the box it is less self-contained than it looks. The quick start asks for OPENAI_API_KEY and TAVILY_API_KEY, which means every query goes to two third parties. Fully local research takes extra configuration of retrievers and an OpenAI-compatible model server. The README also says "unbiased" a lot. Its own disclaimer is more modest: it scrapes many sites and favours the most frequent information, which reduces some errors and repeats popular ones.
It is actively maintained, with releases through 2026 and no published security advisories. Its main competition now is the deep-research features built into commercial assistants, which need no setup. GPT Researcher wins where you need the report generated inside your own application, over your own documents, or with a model you choose.
Where it earns its place
A product engineer at a B2B market-intelligence startup
Generate a sourced competitor brief on demand inside the app
pip install gpt-researcher, call conduct_research() and write_report() from a worker, and store the Markdown. At the README's roughly $0.12 per report, 1,000 briefs a month costs about $120 in model calls plus Tavily search fees.
A policy analyst at a think tank with a folder of PDFs and consultation papers
Write a cited summary of what the documents say about one question
Set DOC_PATH to the folder and choose local as the report source, so the agent researches the files instead of the web. Point OPENAI_BASE_URL at a local model server if the documents must not leave the building.
A solo consultant preparing for a client workshop
Get a five-to-six-page briefing on an unfamiliar industry by tomorrow
Run the Docker Compose stack and use Deep Research mode in the browser UI; the README puts a run at about five minutes and $0.40 with o3-mini. Check citations before quoting, since sources are scraped automatically.
Before you commit
- Setup
- pip install gpt-researcher on Python 3.12+, or clone and run docker compose up --build for the API on port 8000 and the React app on port 3000. Set OPENAI_API_KEY and TAVILY_API_KEY first. Minutes to a first report.
- Integrations
- OpenAI-compatible LLM endpoints, Tavily and other retrievers, MCP servers as retrievers, LangGraph and AG2 multi-agent workflows, LangSmith tracing, Gemini image generation, and an MCP server in the separate gptr-mcp repo.
- Data and privacy
- Queries go to your LLM provider and search API, and pages are scraped directly. Context filtering goes to TypeSafe only if TYPESAFE_API_KEY is set. Local documents stay local unless sent to a hosted model. No telemetry is described in the README, and no GitHub security advisories are published.
- Cost at scale
- No paid edition. By the README's own benchmark, a standard report costs about $0.115 with filtering and $0.192 without; Deep Research is about $0.40 per run with o3-mini on high effort. Search API fees come on top.
- Limits
- Report quality depends on what search returns and what scrapes cleanly. Python 3.12+ only. The benchmark figures are self-reported on 28 tasks.
What’s good
- Python package with a three-line research-and-report API
- Web, local documents and MCP sources in one run
- Low per-report cost by its own benchmark
- Regular releases; no security advisories published
What’s not
- Default setup needs OpenAI and Tavily keys
- Best context filter is a paid third-party API
- Licence differs between GitHub (Apache-2.0) and PyPI (MIT)
- Frequency-based sourcing can repeat popular errors
Best for
Developers building research or report generation into their own product, and analysts who want deep research over their own documents with a model they choose.
GPT Researcher alternatives
GPT Researcher: common questions
- Is GPT Researcher worth it?
- The open deep-research agent with the longest track record, cheap per report and easy to embed, if you accept that its default stack runs on two paid APIs. We rate it 4.1 out of 5.
- How much does GPT Researcher cost?
- Free and open source (Apache-2.0 on GitHub). You pay for the LLM and search APIs it calls.
- What are the downsides of GPT Researcher?
- Default setup needs OpenAI and Tavily keys; best context filter is a paid third-party API; licence differs between GitHub (Apache-2.0) and PyPI (MIT); frequency-based sourcing can repeat popular errors.
- Who is GPT Researcher best for?
- Developers building research or report generation into their own product, and analysts who want deep research over their own documents with a model they choose.
- How much does GPT Researcher cost for a team or in production?
- No paid edition. By the README's own benchmark, a standard report costs about $0.115 with filtering and $0.192 without; Deep Research is about $0.40 per run with o3-mini on high effort. Search API fees come on top.
- Is GPT Researcher safe for company data?
- Queries go to your LLM provider and search API, and pages are scraped directly. Context filtering goes to TypeSafe only if TYPESAFE_API_KEY is set. Local documents stay local unless sent to a hosted model. No telemetry is described in the README, and no GitHub security advisories are published.
- What are the best GPT Researcher alternatives?
- The closest alternatives we have reviewed are smolagents, LangGraph, CrewAI and notebooklm-py.
Answers checked Oct 31, 2026.
Make GPT Researcher? Show this rating on your site
