____ ____ _ _ _
| _ \ ___ ___ ___ __ _ _ __/ ___|| |_ _ _ __| (_) ___
| |_) / _ \/ __|/ _ \/ _` | '__\___ \| __| | | |/ _` | |/ _ \
| _ < __/\__ \ __/ (_| | | ___) | |_| |_| | (_| | | (_) |
|_| \_\___||___/\___|\__,_|_| |____/ \__|\__,_|\__,_|_|\___/
ResearStudio is the first open-source framework for building human-intervenable deep research agents. It enables real-time collaboration between humans and AI, allowing users to pause, edit, and guide AI agents during execution rather than operating in a traditional "fire-and-forget" mode.
- 🤝 Real-time Human-AI Collaboration: Pause, edit, and resume agent execution at any time
- 📊 Multi-Agent Architecture: Hierarchical Planner-Executor system with specialized tools
- 🔧 Comprehensive Tool Suite: Search, code execution, document processing, and more
- 🌐 Web-based Interface: Modern React/Next.js frontend with real-time updates
- 📝 Plan-as-Document: All agent plans are visible and editable in real-time
- 🔒 Sandboxed Execution: Safe, isolated environments for code execution
- 🚀 State-of-the-art Performance: Achieves 70.91% on GAIA benchmark
ResearStudio follows a three-layer architecture:
- L-1 Tool Layer: MCP-based tool suite (search, code, documents, etc.)
- L-2 Agent Core: Planner (GPT-4) and Executor (GPT-4o-mini/o3)
- L-3 Interface Layer: React/Next.js web interface with SSE streaming
┌─────────────────────────────────────────┐
│ L-3: Web Interface │
│ (Next.js + React 19) │
└────────────────┬────────────────────────┘
│ SSE/WebSocket
┌────────────────▼────────────────────────┐
│ L-2: Agent Core │
│ ┌──────────┐ ┌──────────┐ │
│ │ Planner │────▶│ Executor │ │
│ └──────────┘ └──────────┘ │
└────────────────┬────────────────────────┘
│ MCP Protocol
┌────────────────▼────────────────────────┐
│ L-1: Tool Layer │
│ Search │ Code │ Docs │ Video │ Image │
└─────────────────────────────────────────┘
- Python 3.8+
- Node.js 18+
- OpenAI API key
- Git
# Clone the repository
git clone https://github.com/ResearAI/ResearStudio.git
cd ResearStudio
# Run the one-click installer
./setup.sh
# Add your OpenAI API key
nano agent/.env # Add your API key# Install Python dependencies
pip install -r requirements.txt
# Install frontend dependencies
cd frontend
npm install --legacy-peer-deps
cd ..
# Set up environment variables
cp agent/.env.example agent/.env
cp frontend/.env.example frontend/.env.local
# Edit the .env files with your API keys# Start both backend and frontend
./start.sh
# Access the application at http://localhost:3000
# To stop the application
./stop.sh# Terminal 1: Start the backend
cd agent
python app.py
# Terminal 2: Start the frontend
cd frontend
npm run dev
# Access at http://localhost:3000- Installation Guide - Detailed installation instructions
- Developer Guide - Guide for extending and modifying the system
- API Documentation - Complete API reference
- Blog Tutorial - Step-by-step tutorial
- Research Automation: Automate literature reviews and data analysis
- Code Development: AI-assisted programming with human oversight
- Data Processing: Process documents, videos, and images with AI
- Complex Problem Solving: Tackle multi-step problems with human guidance
ResearStudio achieves state-of-the-art results on the GAIA benchmark:
| Level | Score |
|---|---|
| Level-1 | 77.36% |
| Level-2 | 69.77% |
| Level-3 | 61.54% |
| Average | 70.91% |
We welcome contributions! Please see our Contributing Guide for details.
This project is licensed under the MIT License - see the LICENSE file for details.
If you use ResearStudio in your research, please cite:
@article{yang2025researstudio,
title={ResearStudio: A Human-Intervenable Framework for Building Controllable Deep-Research Agents},
author={Yang, Linyi and Weng, Yixuan},
year={2025}
}- OpenAI for GPT models
- The GAIA benchmark team
- All contributors and users
- Email: resear.ai@gmail.com
- Issues: GitHub Issues