GPT Researcher

What it does

Done-for-you service

PartlyYou run the software yourself; managed deployment inside your VPC, custom retrievers and SLAs exist only as an enquiry with no published price.No first-party hosted SaaS today. For managed deployments inside your VPC, custom retrievers connected to private corpora, or SLAs, reach out via…gptr.dev/pricing ↗

How much it does unattended

Runs autonomously

YesPlans subtopics, crawls, validates sources and writes the report end to end in one call; typical deep run is 30-60 seconds.GPT Researcher performs autonomous, deep research - not just search. It intelligently explores and validates multiple sources, focusing only on…docs.gptr.dev/docs/gpt-researcher/mcp-s… ↗

Multi-agent

YesA LangGraph flow runs a seven-agent team - Chief Editor, Researcher, Editor, Reviewer, Revisor, Writer, Publisher - and the agent can also be embedded in someone else's agent team.The research team is made up of 7 AI agentsdocs.gptr.dev/docs/gpt-researcher/multi… ↗

Agent permissions

PartlyConstraints are configuration, not policy: the multi-agent flow's include_human_feedback flag is the only human gate, and with it off the agents run the whole task unattended.include_human_feedback - If true, the user can provide feedback to the agents. If false, the agents will work autonomously.docs.gptr.dev/docs/gpt-researcher/multi… ↗

Which models it runs on

Claude

YesAnthropic is a first-class provider in the supported-LLM list; Claude Sonnet is named as an example model on the pricing page.Thanks to the massive community contributions, GPT Researcher supports over 100 different LLMs such as Gemini, Claude, Llama, etc.gptr.dev ↗

GPT

YesOpenAI is the default provider: the quickstart expects OPENAI_API_KEY and the architecture page names gpt-4o-mini and gpt-4o.Easily connect any LLM of your choice from OpenAI, Anthropic, Groq to Llama 3 and HF.gptr.dev ↗

Open models

YesThe supported-provider list includes ollama, vllm, huggingface, together, deepseek and litellm, so open-weight models run locally or via a gateway.Easily connect any LLM of your choice from OpenAI, Anthropic, Groq to Llama 3 and HF.gptr.dev ↗

Model choice

YesYou choose the model and the retriever; the docs enumerate 22 LLM providers, from openai and anthropic to ollama, vllm and openrouter.The following LLMs are supported by GPTR (though you'll need to install the relevant langchain package separately if you're not using OpenAI).docs.gptr.dev/docs/gpt-researcher/llms/… ↗

Where you use it

In your editor

PartlyNo first-party IDE extension: you register the self-hosted gptr-mcp server in Cursor or Claude Desktop over stdio.Supports stdio (Claude Desktop, Cursor), SSE (Docker, n8n, web), and Streamable HTTP.gptr.dev/llms.txt ↗

On the command line

Yescli.py takes a query plus --report_type and --tone and writes the report to the outputs directory.This command-line interface (CLI) tool allows you to generate research reports using the GPTResearcher class.docs.gptr.dev/docs/gpt-researcher/getti… ↗

In a browser

PartlyA React/Next.js frontend ships with the project and runs at localhost:3000 beside the Python server; gptr.dev itself hosts no product.Visit localhost:3000 on any browser and enjoy researching!docs.gptr.dev/docs/gpt-researcher/getti… ↗

Whose machine it runs on

Self-hosted

YesSelf-hosting is the only deployment: pip package, FastAPI server, or docker-compose on your own machine or container platform.Self-host on your own machine, server, or container.gptr.dev/pricing ↗

Open source

YesMIT-licensed, with the core agent and the gptr-mcp server both published as source.GPT Researcher is published under the MIT License.gptr.dev/pricing ↗

What it costs to run

How it meters

YesNo vendor billing at all: you meter per research run against your own LLM and retriever providers, and the vendor publishes a per-run ballpark.Indicative ballpark for a single deep-research run with default OpenAI + Tavily settings: a few cents to a few dollars of provider spend.gptr.dev/pricing ↗

Free tier

YesNo signup, no call and no gateway: install the package and supply your own provider keys.There is no SaaS subscription, no API gateway, and no rate limit on this site.gptr.dev/pricing ↗

API access

YesThe self-hosted FastAPI server exposes REST plus a /ws WebSocket, documented by an OpenAPI 3.1 spec; there is no vendor-hosted endpoint to call.The server exposes a WebSocket endpoint at /ws for streaming research, plus a REST surface for kicking off research jobs and retrieving reports.gptr.dev/llms-full.txt ↗

MCP server

YesThe official gptr-mcp server exposes deep_research, quick_search, write_report, get_research_sources and get_research_context over stdio, SSE or Streamable HTTP.The GPT Researcher MCP Server provides Model Context Protocol (MCP) integration for GPT Researcher, allowing AI assistants to perform autonomous,…docs.gptr.dev/docs/gpt-researcher/mcp-s… ↗

Bring your own key

YesBring-your-own-keys is the only model: you supply the LLM and retriever keys at runtime and can point it at a local model instead.You self-host the agent and pay only your underlying LLM and search-API providers.gptr.dev/pricing ↗

Buying it for a team

A company can buy it

PartlyNothing is sold: the vendor says it works with enterprises, but the only commercial route is an unpriced enquiry for managed deployment or an SLA.We work closely with individuals, startups and enterprises to provide the best research experience for your business or AI automation.gptr.dev ↗

Seat model

NoOne MIT-licensed tier priced at $0 for the whole product; no seats, user counts or subscription anywhere on the pricing page.There is no SaaS subscription, no API gateway, and no rate limit on this site.gptr.dev/pricing ↗

Pooled budget

NoThere is no vendor-side balance; spend sits entirely with the LLM and retriever accounts you own.None imposed by GPT Researcher. Limits come from your underlying LLM / retriever providers' rate limits.gptr.dev/pricing.md ↗

Admin controls

PartlyConstraints are configuration, not an admin plane: whoever deploys it caps iterations, subtopics and results per query, and picks the retrievers.Research runs are bounded by max_iterations, max_subtopics, and max_search_results_per_query configs.gptr.dev/llms-full.txt ↗

Audit log

NoThere is no account or admin plane whose actions could be recorded.There is no GPT Researcher account system.gptr.dev/llms-full.txt ↗

Single sign-on

NoNo SAML or OIDC, because there is nothing to log into: the only credentials are the provider API keys you supply at runtime.There is no GPT Researcher account system.gptr.dev/llms-full.txt ↗

What happens to your code

Opt out of training

YesNothing reaches the vendor; data leaves only through the providers you choose, and a local model with SearXNG keeps every query on your own hardware.Only via the LLM and retriever providers you configure. Self-host with a local model and SearXNG to keep everything on-prem.gptr.dev/pricing ↗

Data residency

PartlyResidency is wherever you deploy: the vendor holds no data and publishes no region list, and an on-prem configuration is documented.Only via the LLM and retriever providers you configure. Self-host with a local model and SearXNG to keep everything on-prem.gptr.dev/pricing ↗

Getting out

YesReports export to PDF, Word, Markdown, JSON and CSV from your own deployment, and there is no subscription, so no credit balance can expire.Export your research to any format from PDF, Word, Markdown to JSON and CSV.gptr.dev ↗

Not established: Certifications