lasyaMundrathi/mcp-deep-research-assistant

Deep Researcher: Multi-Agent Research Assistant (Streamlit & MCP)

A modular, AI-powered research assistant that performs in-depth web research and summarizes findings with citations. Built with CrewAI, enhanced by Linkup for real-time search, and accessible via Streamlit or as an MCP tool for IDEs (e.g., Cursor).

🏗️ System Architecture

Key Components

Agents (CrewAI)

• Web Search Agent: Fetches real-time internet results.
• Research Analyst Agent: Synthesizes, deduplicates, and fact-checks findings.
• Technical Writer Agent: Composes a well-cited article.

Tools

• Linkup Search Tool: Interfaces with Linkup API for reliable, real-time search.

User Interfaces

• Streamlit UI: Query and view responses in-browser.
• MCP Server: Call the researcher from any IDE supporting MCP (e.g., Cursor).

How It Works

1. User enters a query (Streamlit or MCP).

2. CrewAI workflow:

   • Web Search Agent collects sources
   • Analyst Agent reviews/verifies
   • Writer Agent composes the result

3. Final response shown in UI or returned to IDE.

Setup

Prerequisites

• Python 3.11+
• Linkup API Key

Run Streamlit App

streamlit run app/streamlit_app.py

Run as MCP Server (for IDEs like Cursor)

python server.py

Add to your mcp.json:

{
  "mcpServers": {
    "crew_research": {
      "command": "uv",
      "args": ["--directory", "./Multi-Agent-deep-researcher-mcp-windows-linux", "run", "server.py"],
      "env": {"LINKUP_API_KEY": "your_linkup_api_key_here"}
    }
  }
}

Folder Structure

deep-researcher/
├── agents.py/
│   ├── web_search_agent
│   ├── research_analyst
│   └── technical_writer
├── app.py
├── server.py
├── pyproject.toml
└── README.md

Credits

• CrewAI
• Linkup API
• MCP Client from Cursor