OpenDeepSearch
An open-source alternative to Perplexity Deep Research using the Model Context Protocol (MCP).
Overview
OpenDeepSearch is a powerful research tool that performs comprehensive, in-depth research on complex topics. It combines the structured thinking approach of Sequential Thinking with the search capabilities of Brave Search to provide detailed, well-sourced research reports.
Features
- Comprehensive Research: Breaks down complex questions into manageable sub-questions
- Iterative Search: Performs multiple searches to gather diverse information
- Intelligent Analysis: Analyzes search results to extract relevant information
- Synthesis: Combines findings into a coherent, well-structured report
- Citations: Includes sources for all information in the report
- MCP Integration: Seamlessly integrates with Claude Desktop, Cursor, and other MCP clients
- WebSockets: Supports integration with Smithery and other MCP clients
- Publication: Allows publishing the research tool on the Smithery platform for easy access
Installation
Prerequisites
NPM Installation
npm install -g open-deep-research
Running with NPX
BRAVE_API_KEY=your_api_key npx open-deep-research
Local Installation
# Clone the repository
git clone https://github.com/tositon/open-deep-research.git
cd open-deep-research
# Install dependencies
npm install
# Build the project
npm run build
# Run with Brave Search API
BRAVE_API_KEY=your_api_key npm start
Installation via Smithery
# Install for Claude
npx @smithery/cli install open-deep-research --client claude
# Install for Cursor
npx @smithery/cli install open-deep-research --client cursor
When installing via Smithery, you will be prompted to enter a Brave Search API key.
Usage
With Claude Desktop
Add the following to your Claude Desktop configuration:
{
"mcpServers": {
"open-deep-research": {
"command": "npx",
"args": [
"-y",
"open-deep-research"
],
"env": {
"BRAVE_API_KEY": "your_api_key_here"
}
}
}
}
With Cursor
In Cursor, you can add the MCP server with:
claude mcp add "open-deep-research" npx open-deep-research
Make sure to set the BRAVE_API_KEY environment variable before running Cursor.
Example Queries
- "What are the latest developments in quantum computing?"
- "Compare and contrast different approaches to climate change mitigation"
- "Explain the history and impact of the Renaissance on European art"
- "What are the pros and cons of different renewable energy sources?"
How It Works
- Question Analysis: The system analyzes the main question and breaks it down into sub-questions
- Iterative Search: For each sub-question, the system performs searches using Brave Search API
- Result Analysis: The system analyzes the search results to extract relevant information
- Synthesis: The system combines the findings into a coherent report
- Citation: All information is properly cited with sources
Development
Setup
git clone https://github.com/tositon/open-deep-research.git
cd open-deep-research
npm install
Build
Run in Development Mode
BRAVE_API_KEY=your_api_key npm run dev
Testing
Testing with MCP Inspector
Для тестирования MCP сервера можно использовать MCP Inspector, который предоставляет удобный интерфейс для взаимодействия с инструментами:
# Установка и запуск MCP Inspector
npx @modelcontextprotocol/inspector
# Запуск сервера в другом терминале
BRAVE_API_KEY=your_api_key npm start
После запуска Inspector, откройте браузер и перейдите по адресу http://localhost:5173. Подключитесь к WebSocket серверу, используя URL ws://localhost:3000.
Примеры запросов для тестирования инструментов
В интерфейсе MCP Inspector вы можете выбрать инструмент и настроить параметры запроса:
Тестирование Brave Web Search
{
"query": "latest quantum computing advancements",
"count": 5
}
Тестирование Sequential Thinking
{
"thought": "Начинаю анализ проблемы глобального потепления",
"thoughtNumber": 1,
"totalThoughts": 5,
"nextThoughtNeeded": true
}
Тестирование Deep Research
{
"query": "Сравнение различных источников возобновляемой энергии",
"action": "start",
"maxSubQuestions": 3
}
Testing with Claude or Cursor
После установки сервера через Smithery или локально, вы можете использовать его с Claude Desktop или Cursor, выбрав соответствующий MCP сервер в настройках.
Publishing on Smithery
To publish the server on the Smithery platform:
- Ensure the repository is hosted on GitHub and is public
- Register on the Smithery platform
- Authenticate via GitHub to connect with the repository
- Go to the "Deployments" tab on the server page
- Click the "Deploy on Smithery" button
- Follow the deployment setup instructions
After publishing, users can install the server using the Smithery CLI:
npx @smithery/cli install open-deep-research --client claude
Contributing
Contributions are welcome! Please feel free to submit a Pull Request.
License
This project is licensed under the MIT License - see the LICENSE file for details.
Acknowledgments
- Inspired by Perplexity Deep Research
- Built on the Model Context Protocol
- Uses Sequential Thinking approach for structured research
- Powered by Brave Search API