Pricing
The software is open source and free to install and self host. Costs come from the LLM and search provider API keys you connect, which are billed by those providers. No paid plans are described on the homepage.
Learn moreGPT Researcher
Open source autonomous research agent that writes cited reports
Advanced API
Own GPT Researcher? Verify ownership to get a dofollow link
What is GPT Researcher?
GPT Researcher is an AI search tool. An open source research agent that plans a question, runs parallel researchers across web and local sources, filters passages, and writes long reports with inline citations. Configure your own LLM and search engine, export to PDF, Word, Markdown, JSON or CSV, or call it from an MCP client.
GPT Researcher is an open source autonomous research agent. You give it a question, and it writes a plan, splits that plan into a set of narrower research questions, then runs a separate researcher on each question in parallel. Sources are scraped and filtered, only the passages that matter are passed forward, and a writer assembles the result into a long report with inline citations. Every claim in the output traces back to a source, which is the core promise of the project. The architecture is deliberately modular. A planner handles decomposition, parallel researchers pull from more than twenty retrievers (Tavily, Google, Bing, Exa, arXiv, PubMed and others, plus any MCP server), and a context filter (Jev by default, with keyword ranking or embeddings as alternatives) trims noise before anything reaches the model. Deep research mode feeds findings back in as new questions, so a run grows a tree of subtopics rather than making a single pass. You choose the LLM and the search engine, and reports export to Markdown, PDF, Word, JSON or CSV. The tool is built for people who already think in code. Installation is a pip command on Python 3.11 or later, and the documented usage is a short async script. That makes it a natural fit for developers, research engineers, data teams and analysts who want research inside their own pipelines, and for anyone wiring a research agent into Claude, Cursor or n8n through the companion MCP server. Local document research and hybrid web plus local runs extend it to private material. Against alternatives, GPT Researcher is not a polished consumer chat product. It competes on transparency, configurability and citation rigor, and it points to Carnegie Mellon's DeepResearchGym benchmark (May 2025), where it ranked first on citation quality, report quality and coverage ahead of Perplexity and OpenAI systems. Teams that want a hosted interface and zero setup will find the setup cost real; teams that want an auditable, self-managed research pipeline get a mature, widely adopted option.
How do you use GPT Researcher?
1Read the introduction and prerequisites
Start with the getting started introduction to confirm requirements (Python 3.11 or later) and understand the planner, researcher and writer flow before installing.
2Install the package
Run pip install gpt-researcher, then add your LLM and search provider API keys to the environment so the agent can call a model and a retriever.3Run your first research query
Create a GPTResearcher instance with your question, await conduct_research(), then await write_report() to get a cited report back as markdown or another format.4Configure retrievers and context filter
Set RETRIEVER to combine sources such as tavily and mcp, and choose CONTEXT_FILTER to control which passages survive filtering before the writer sees them.5Enable deep research for broader coverage
Turn on deep research and set depth and iteration values so findings generate follow up questions and the run explores subtopics rather than stopping after one pass.6Expose research to an MCP client
Clone the gptr-mcp repository, register it in your MCP client config with the required API keys, and call deep_research, quick_search or write_report from Claude, Cursor or n8n.
Pros and cons
Pros
- Every claim in the report carries a source citation, which makes outputs auditableAI
- Planner plus parallel researchers covers subtopics in one run instead of one query at a timeAI
- Deep research mode expands findings into new questions by breadth and depthAI
- Any LLM and any search engine can be swapped in, with 20+ retrievers documentedAI
- Reports exceed typical token limits (2K+ words) and export to PDF, Word, Markdown, JSON and CSVAI
Cons
- Requires Python 3.11+ and code to run; there is no hosted point and click interface on the siteAI
- You supply and pay for your own LLM and search API keys, so running costs are on youAI
- Local keyword ranking is the fallback filter when no TypeSafe key is set, which is weaker than the defaultAI
- Open source project with a fast moving setup; no formal support channel is describedAI
- Quality depends heavily on retriever choice and filter configuration, so defaults need tuningAI
How much does GPT Researcher cost?
Support
Support runs through the documentation site, the GitHub repository and a community Discord server. The homepage does not describe a paid or dedicated support tier.
Integrations
Works with any LLM (OpenAI, Anthropic, Groq, Llama 3, Hugging Face) and major search engines. The gptr-mcp server exposes research tools to Claude, Cursor, n8n and any MCP client.
Learn moreFeatures
Planner driven question decomposition, parallel researchers, 20+ retrievers including Tavily, Google, Bing, Exa, arXiv and PubMed, MCP server support, research over local documents with hybrid web plus local runs, Jev, BM25 or embedding based context filtering, deep research looping by depth and breadth, long report generation, multi agent support, and export to PDF, Word, Markdown, JSON and CSV.
Learn moreFrequently asked questions about GPT Researcher
How much does GPT Researcher cost?
The software is open source and free to install and self host. Costs come from the LLM and search provider API keys you connect, which are billed by those providers. No paid plans are described on the homepage.Does GPT Researcher have an API?
Yes, GPT Researcher offers an API.How do you use GPT Researcher?
The walkthrough on this page covers 6 steps: 1. Read the introduction and prerequisites 2. Install the package 3. Run your first research query 4. Configure retrievers and context filter 5. Enable deep research for broader coverage 6. Expose research to an MCP client.What platforms does GPT Researcher support?
GPT Researcher is available on Web App, Windows, MacOS and Linux.What does GPT Researcher integrate with?
Works with any LLM (OpenAI, Anthropic, Groq, Llama 3, Hugging Face) and major search engines. The gptr-mcp server exposes research tools to Claude, Cursor, n8n and any MCP client.What are the limitations of GPT Researcher?
Requires Python 3.11+ and code to run; there is no hosted point and click interface on the site. You supply and pay for your own LLM and search API keys, so running costs are on you. Local keyword ranking is the fallback filter when no TypeSafe key is set, which is weaker than the default.
Status
StatusActive
Views0
Outbound clicks0
Added10/21/2026
Platforms
Web AppWindowsMacOSLinux
Pricing
FreePaid
