Are you looking for a robust, open-source alternative to proprietary research tools like OpenAI Deep Research or Google's Deep Researcher? In this guide, you'll learn how to set up your own deep research agent using Model Context Protocol (MCP) servers such as Sequential-Thinking and Exa. This solution gives API developers and technical teams flexibility, transparency, and full control over their research workflows.
💡 Supercharge your API workflow with Apidog:
Design, test, mock, and document APIs in a seamless, all-in-one platform. Integrate your open-source research workflows and experience efficient API development with Apidog's design tools, testing suite, mocking features, and documentation support.
Why Build an Open-Source Deep Research Agent?
Open-source research agents empower API engineers and technical teams to automate information gathering and analysis while retaining control over data and configurations. Compared to closed-source platforms, you can:
- Integrate your preferred AI models and tools
- Ensure transparency in data processing
- Customize workflows for your team's needs
- Avoid vendor lock-in and reduce costs
With MCP servers like Sequential-Thinking (for structured reasoning) and Exa (for AI-powered web search), you can assemble a highly capable research environment tailored to your use case.
💡 Explore more at HiMCP:
Discover over 1,682 MCP servers and clients to boost your AI development workflow. See the full directory here: HiMCP.ai - Discover 1682+ MCP Servers
What Is an Open-Source Deep Research Agent?
An open-source deep research agent is a programmable tool that automates complex research by combining AI-driven reasoning with web search capabilities through the Model Context Protocol. Here’s how it works:
- Connects to MCP servers (like Sequential-Thinking or Exa) for intelligent query processing
- Integrates with LLMs (e.g., DeepSeek V3, Claude Sonnet, GPT-4o) for summarizing and interpreting data
- Fetches and analyzes web data for up-to-date, comprehensive research
- Open-source architecture ensures control, transparency, and customization
Who Should Use This Agent?
This workflow is ideal for:
- Researchers & Academics: Accelerate literature reviews and knowledge synthesis
- Journalists & Writers: Automate background research and fact-checking
- Backend/API Developers: Build and test custom research features or integrate into apps
- Analysts & Policy Makers: Extract insights from large datasets and reports
- Students & Lifelong Learners: Efficiently summarize and understand complex concepts
When Does an Open-Source Research Agent Add Value?
- Handling large-scale or repetitive research tasks: Automate multi-source data collection and analysis
- Custom AI model integration: Use your preferred LLMs for domain-specific needs
- Transparency and control: Gain full visibility into how data is processed
- Boosting productivity: Accelerate workflows by combining sequential reasoning and web search
Step-by-Step: Setting Up Your Deep Research Agent
Prerequisites
- Windsurf IDE: Download the latest version from the official site
- Node.js: v20 or higher
- npm: v7 or higher (latest recommended)
1. Create and Open Your Project
- Create a folder (e.g.,
deep_researcher) - Open it in Windsurf IDE, which supports MCP server integration
2. Install Sequential-Thinking MCP Server
Run the following command to install Sequential-Thinking MCP:
npx -y @smithery/cli@latest install @smithery-ai/server-sequential-thinking --client windsurf --config "{}"
Verify configuration:
Check your Windsurf mcp_config.json (typically found at C:/Users/You/.codeium/windsurf/mcp_config.json). It should look like this:
{
"mcpServers": {
"server-sequential-thinking": {
"command": "cmd",
"args": [
"/c",
"npx",
"-y",
"@smithery/cli@latest",
"run",
"@smithery-ai/server-sequential-thinking",
"--config",
"\"{}\""
]
}
}
}
If empty, copy the example above or check the official GitHub repository.
Test the server:
Try a sample prompt:
>> Use sequential thinking to help me develop a simple Flappy Bird Python game.
3. Set Up Exa Web Search MCP Server
a. Get an Exa API Key
- Visit the Exa website, create an account, and generate a free API key from your profile.
b. Clone and Build the Exa MCP Server
git clone https://github.com/exa-labs/exa-mcp-server.git
cd exa-mcp-server
npm install
npm run build
npm link
c. Configure Exa MCP in Windsurf
Update your mcp_config.json:
{
"mcpServers": {
"server-sequential-thinking": {
"command": "cmd",
"args": [
"/c",
"npx",
"-y",
"@smithery/cli@latest",
"run",
"@smithery-ai/server-sequential-thinking",
"--config",
"\"{}\""
]
},
"exa": {
"command": "npx",
"args": ["C:/Research_agent/exa-mcp-server/build/index.js"],
"env": {
"EXA_API_KEY": "YOUR_API_KEY"
}
}
}
}
Replace "YOUR_API_KEY" with your actual key.
Test Exa server:
Sample prompt:
>> Find blog posts about AGI.
Using Your Deep Research Agent: Practical Example
With Sequential-Thinking and Exa MCP servers set up, try this advanced prompt:
>> When using sequential-thinking, use as many steps as possible. Each step of sequential-thinking must use branches, isRevision, and needsMoreThoughts, and each branch must have at least 3 steps. Before each step of sequential-thinking, use Exa to search for 3 related web pages and then think about the content of the web pages. The final reply should be long enough and well-structured. Question: What is metaphysics?
This will generate a detailed, stepwise answer backed by real web searches.
Sample web search output:
Sample agent output:
Key Features and Benefits
- Full control and privacy: Open-source MCP servers keep your data in your hands
- Custom AI model support: Integrate DeepSeek v3, Claude Sonnet, GPT-4o, or others
- Flexible workflows: Tailor the research process to your team's needs
- Cost-effective: Reduce or eliminate dependency on expensive proprietary tools
For a direct comparison:
Ollama Deep Research, the Open-Source Alternative to OpenAI Deep Researcher. This guide covers setup, features, pricing, and why it’s a better choice.
Troubleshooting Common Issues
- MCP Server Not Configured: Double-check your
mcp_config.jsonand ensure correct API keys and paths. - API Key Problems: Make sure your keys are correct, active, and have not exceeded usage limits.
- Model Not Responding: Verify installation and update dependencies as needed.
- Web Search Issues: Ensure internet connectivity and valid API configurations for Exa.
Conclusion
Building an open-source deep research agent with MCP servers gives API teams unprecedented flexibility, control, and efficiency for advanced research and development workflows. Integrate these tools with your existing API processes in Apidog to further streamline design, testing, and documentation in a single, developer-friendly environment.
