ResearchOS: An Autonomous AI Research Engine Based on Multi-Agent Workflow

ResearchOS is an open-source autonomous research assistant that automates search, reading, writing, and review through a LangChain multi-agent pipeline, providing an end-to-end intelligent research solution for knowledge workers. Its core idea is "let agents do research, let humans make decisions"—users only need to input a research topic, and the system automatically generates structured research reports.

In the era of information explosion, traditional research processes require repeated switching between search engines, literature databases, and note-taking tools, with manual completion of tedious tasks like information collection, filtering, integration, and writing. ResearchOS is designed to address these pain points.

ResearchOS uses a LangChain-based pipeline with four specialized agents:

1. Search Agent: Converts topics into queries and retrieves high-quality sources via Tavily API.
2. Reader Agent: Extracts full text from pages using BeautifulSoup4.
3. Writer Agent: Integrates content to generate coherent reports with Mistral AI.
4. Critic Agent: Evaluates report accuracy, completeness, and logic to reduce hallucinations. Execution flow: User input → Search → Read → Write → Critic → Final report. Each agent uses Groq/Mistral for fast reasoning, enabling parallel execution.

Key tech stack choices:

• LangChain: Mature LLM framework with modular design for agent orchestration.
• Mistral AI: Balances cost and performance, accelerated by Groq for fast token generation.
• Tavily Search API: AI-optimized, returns RAG-friendly formatted results.
• BeautifulSoup4: Flexible HTML parsing for content extraction.
• Streamlit: Rapid interactive UI development with real-time workflow visualization.

1. Real-time workflow visualization: Streamlit interface shows current stage (searching, reading, etc.) and progress.
2. Structured report generation: Produces complete Markdown reports with clear structure and in-depth analysis.
3. Built-in quality review: Critic agent ensures output reliability.
4. Export function: Supports Markdown download for further editing or archiving.

ResearchOS applies to:

• Academic literature research: Quickly grasp field status, main schools, and key results.
• Industry trend analysis: Generate preliminary reports on industry dynamics and competition.
• Tech selection research: Understand pros/cons, community activity, and use cases of candidate technologies.
• News event tracking: Summarize background and development脉络 in minutes. It saves time by automating tedious research tasks.

Current limitations: No dialogue memory, incomplete citation tracking, only Markdown output, no vector database integration. Future plans: Memory-enhanced agents, citation tracking, PDF export, AI reasoning loop, vector database integration, recursive subqueries, configurable agent roles and report formats.

ResearchOS is open-sourced under the MIT license. Initiated by Maulik Gupta, the code is hosted on GitHub. Community contributions are welcome via issues and pull requests.