assafelovic/gpt-researcher

What it solves

GPT Researcher is designed to automate the process of deep research, replacing the manual, time-consuming task of browsing multiple websites and synthesizing information. It addresses common LLM limitations such as hallucinations, outdated training data, token limits for long-form reports, and the bias or shallowness resulting from limited web sources.

How it works

The system uses a multi-agent architecture consisting of a planner and execution agents. A planner generates specific research questions based on a query, while execution agents (crawler agents) gather information from the web or local documents. A publisher then aggregates these findings, summarizes them, and tracks sources to produce a factual, unbiased research report with citations.

Who it’s for

It is intended for individuals and organizations who need detailed, factual, and unbiased research reports on any given task, as well as developers looking to integrate deep research capabilities into their own AI assistants via PIP or MCP integration.

Highlights

• Comprehensive Reports: Generates detailed research reports exceeding 2,000 words using over 20 sources.
• Hybrid Research: Supports both web-based research and research on local documents (PDF, CSV, Excel, Word, etc.).
• Deep Research Mode: Employs a recursive, tree-like exploration pattern to dive deeper into subtopics.
• AI Visuals: Automatically generates and embeds inline illustrations using Google Gemini.
• Multi-Agent Frameworks: Integrates with LangGraph and AG2 for specialized agentic workflows.
• Extensible Integration: Available as a PIP package, Docker image, or Claude Skill, and supports the Model Context Protocol (MCP) for connecting to external data sources like GitHub.