MCP-Deep-Researcher

A sophisticated multi-agent research system that breaks down broad queries into focused research questions, searches the web comprehensively, and synthesizes findings into detailed markdown reports.

Perfect for: Research tasks, competitive analysis, trend research, fact-gathering, and knowledge synthesis across multiple sources.

• [LangGraph](https://github.com/langchain-ai/langgraph) — Graph-based agent orchestration with typed state
• [OpenAI GPT-4o](https://platform.openai.com/) — LLM for planning and synthesizing research
• [Tavily](https://tavily.com/) — Advanced web search with source citations
• [Streamlit](https://streamlit.io/) — Interactive web UI for research queries
• [MCP](https://modelcontextprotocol.io/) — Model Context Protocol for Claude/Cursor integration

The system uses a 3-node LangGraph pipeline:

User Query → Planner → Searcher → Synthesizer → Markdown Report

1. Planner Node — Decomposes the broad query into 3-5 targeted research questions using GPT-4o with structured output
2. Searcher Node — Executes web searches for each question using Tavily API, collecting results with source URLs
3. Synthesizer Node — Analyzes search results and creates a comprehensive markdown report with findings, knowledge gaps, and citations

Output is clean, well-structured markdown with:

• Executive summary
• Key findings per research question
• Identified knowledge gaps
• Properly cited sources

You'll need API keys for:

• OpenAI (GPT-4o): Get from platform.openai.com/api-keys
• Tavily (Web search): Get from app.tavily.com/home

1. Clone the repository:

git clone https://github.com/sid12super/MCP-Deep-Researcher.git
cd MCP-Deep-Researcher

1. Install dependen