research-mcp

karashiiro/research-mcp

3.2

If you are the rightful owner of research-mcp and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

Deep research MCP server is a work-in-progress project focused on developing a robust model context protocol server.

(WIP) Deep Research MCP Server

A research orchestration MCP server. This system implements a hierarchical research architecture where a lead researcher coordinates multiple specialized research agents to conduct investigations on complex topics, generating detailed reports with proper citations and analysis.

Modeled loosely after Claude's Research Tool.

Quick Start

Prerequisites

  • Python 3.12 or higher
  • uv package manager
  • Brave Search API key
  • AWS Bedrock access (recommended) or Ollama setup

Installation

  1. Clone the repository:

    git clone <repository-url>
cd research-mcp

  2. Install dependencies:

    uv sync

  3. Configure environment:

    cp .env.example .env
# Edit .env with your API keys and model settings

Environment Variables

# Required: Brave Search API
BRAVE_API_KEY=your_brave_api_key_here

# Model Configuration
MODEL_TYPE=bedrock  # or "ollama"

# For Bedrock (recommended)
BEDROCK_MODEL=anthropic.claude-3-5-sonnet-20241022-v2:0
BEDROCK_REGION=us-east-1

# For Ollama (alternative)
OLLAMA_HOST=http://localhost:11434
OLLAMA_MODEL=your_model_name

Usage

MCP Server

Start the MCP server:

uv run mcp dev src/mcp_server/server.py

Use via MCP client: The server provides a conduct_research tool that accepts a topic string and returns a research report.

Using Docker (with Bedrock backend)

Create a .env file with appropriate environment variables, and then Add the following command to your MCP configuration:

docker run -i --rm -v /path/to/your.env:/app/.env:ro -v /home/you/.aws:/home/mcp/.aws ghcr.io/karashiiro/research-mcp:main

Advanced Usage

Clear cache (for fresh web searches):

rm -rf cache/

View research progress logs:

cat logs/research_results.log

Architecture

sequenceDiagram
    participant User as User
    participant Orchestrator as Research Orchestrator
    participant Lead as Lead Researcher
    participant ResearchTeam as Research Team
    participant Web as Web Search
    participant Reviewer as Citation Reviewer

    User->>Orchestrator: "Research this topic"

    Orchestrator->>Lead: Begin research
    Note over Lead: Breaks topic into subtopics

    Lead->>ResearchTeam: "Research these subtopics"

    par Parallel Research
        loop for each subtopic
            ResearchTeam->>Web: Search for current data
            Web-->>ResearchTeam: Latest information
        end
    end

    Note over ResearchTeam: Consolidates findings internally
    ResearchTeam-->>Lead: Initial research package

    opt follow-up research needed
        Note over Lead: Reviews findings, identifies deeper investigation areas
        Lead->>ResearchTeam: "Research these follow-up topics"

        par Follow-up Research
            loop for follow-up topics
                ResearchTeam->>Web: Search for additional depth
                Web-->>ResearchTeam: Detailed information
            end
        end

        Note over ResearchTeam: Consolidates follow-up findings
        ResearchTeam-->>Lead: Additional research package
    end

    Lead->>Lead: Draft final research report

    Lead->>Reviewer: "Check citations"
    Reviewer-->>Lead: Citation feedback

    Lead->>Lead: Revise final research report
    Lead-->>Orchestrator: Final research report

    Orchestrator-->>User: Comprehensive research with citations