GoMCPTest: Proof of Concept for MCP with Custom Host

Overview: What is GoMCPTest?

GoMCPTest is a proof-of-concept (POC) project that demonstrates how to implement a Model Context Protocol (MCP) server using a custom-built host. The goal is to enable easy testing and iteration of agentic systems, making it straightforward for developers to create specialized agents tailored to different use cases such as security vulnerability scanning, code review analysis, data processing, and automated documentation generation.

🔧 Core Features & MCP Capabilities

GoMCPTest includes several key features that align with the Model Context Protocol (MCP):

1. OpenAI Compatibility: The API is designed to be compatible with the OpenAI v1 chat completion format.
2. Google Gemini Integration: This integration utilizes the Vertex AI API for interactions with Google Gemini models, providing advanced natural language processing and reasoning capabilities.
3. Streaming Support: The server supports streaming responses, enabling real-time updates in conversations.
4. Function Calling: Allows Gemini to call external functions, incorporating their results into chat responses seamlessly.
5. MCP Server Interaction: Demonstrates interaction with a hypothetical MCP (Model Control Plane) server for tool execution and data management.

Real-world AI Workflow Scenarios

1. Security Vulnerability Scanning Agent:
   • This agent can be specialized to scan codebases for security vulnerabilities using custom command-line tools or API integrations.
2. Code Review Analysis Agent:
   • Able to analyze pull requests for potential issues by parsing and processing the content, integrating with various Git repositories.

MCP Protocol Implementation Details

The protocol flow is designed to ensure seamless communication between the AI application (MCP client) and the server (MCP host). Here’s a breakdown of the typical interaction:

graph TD
    A[AI Application] -->|MCP Client| B[MCP Protocol]
    B --> C[MCP Server]
    C --> D[Data Source/Tool]
    style A fill:#e1f5fe
    style C fill:#f3e5f5
    style D fill:#e8f5e8

This diagram showcases the flow from the AI application initiating a request via an MCP client, through to the protocol handling the interaction with the server and any associated data resources.

⚙️ MCP Architecture & Protocol Implementation

GoMCPTest's architecture is modular and composed of several key components:

1. Host/openaiserver: Implements a custom host that mimics the OpenAI API, using Google Gemini for advanced natural language processing.
2. Tools Directory (tools):
   • Bash: Execute bash commands
   • Edit: Edit file contents
   • GlobTool: Find files matching glob patterns
   • GrepTool: Search file contents with regular expressions
   • LS: List directory contents
   • Replace: Replace entire file contents
   • View: View file contents

These components work together to provide a fully functional MCP server, enabling various AI agents and tools.

🚀 Getting Started with Installation

To get started with GoMCPTest:

1. Prerequisites:

   • Ensure you have installed and configured Go >= 1.21.
   • Access to the Vertex AI API on Google Cloud Platform.
   • The tools github.com/mark3labs/mcp-go are required.

2. Bash Installation:

   make all # or individual tool installation
   

3. Environment Configuration: Read the .envrc file in the bin directory to set up the necessary environment variables:

   export GCP_PROJECT=your-project-id
   export GCP_REGION=your-region
   export GEMINI_MODELS=gemini-2.0-flash
   export IMAGEN_MODELS=imagen-3.0-generate-002
   export IMAGE_DIR=/tmp/images
   

4. Running the Server:

   • Navigate to the host/openaiserver directory.

   • Set environment variables as required.

   • Run the server with:

     go run .
     

💡 Key Use Cases in AI Workflows

With GoMCPTest, developers can leverage a variety of AI applications by integrating them into workflows. Here are some examples:

Use Case 1: Automated Code Analysis Agent

• Objective: Analyze codebases for security vulnerabilities.
• Implementation: The dispatch_agent could be specialized to scan codebases using custom scripts or tools integrated through the GoMCPTest API.

Use Case 2: Data Visualization Agent

• Objective: Process and visualize complex datasets.
• Implementation: An agentic system can pull data from a database, process it with machine learning models via GoMCPTest, and then generate visualizations for analysis.

🔌 Integration with MCP Clients

GoMCPTest is designed to be compatible with several popular MCP clients:

MCP Client	Resources	Tools	Prompts	Status
Claude Desktop	✅	✅	✅	Full Support
Continue	✅	✅	✅	Full Support
Cursor	❌	✅	❌	Tools Only

This matrix indicates that while other clients support tools, full compatibility with resources (like custom databases or storage systems) is not yet available.

📊 Performance & Compatibility Matrix

MCP Configuration Code Sample

Here’s a sample configuration snippet demonstrating how to set up an MCP server:

{
  "mcpServers": {
    "[server-name]": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-[name]"],
      "env": {
        "API_KEY": "your-api-key"
      }
    }
  }
}

This JSON snippet configures the server settings, including the command to run and environment variables.

🛠️ Advanced Configuration & Security

To ensure robust security and performance:

1. Environment Variables: Properly configure GCP_PROJECT, GEMINI_MODELS, IMAGEN_MODELS, and other required parameters.
2. Chroot Environment: Preferably use a chroot or container environment to prevent potential system damage from tools that have the ability to execute commands and modify files.

❓ Frequently Asked Questions (FAQ)

1. Q: How do I configure GoMCPTest for full compatibility with multiple clients?
   • A: Ensure you set up the required environment variables, including API_KEY for different MCP clients.
2. Q: Can GoMCPTest be integrated with custom databases or other resources?
   • A: Currently, only limited integration with such resources is supported; full compatibility requires additional development effort.
3. Q: What are the limitations of GoMCPTest as a POC?
   • A: While GoMCPTest provides a solid foundation, it currently lacks full support for all MCP clients and resource integrations. Further development will be required to address these gaps.
4. Q: How do I run GoMCPTest in a production environment?
   • A: Ensure proper security configurations and consider running in a containerized or chrooted environment with minimal system access rights.
5. Q: Are there any known compatibility issues between GoMCPTest and specific MCP clients?
   • A: Refer to the compatibility matrix for detailed information on supported features; some clients may have limited support.

👨‍💻 Development & Contribution Guidelines

Contributions are welcome and can be made by:

1. Forking the Repository:
   • Clone or fork this repository from GitHub.
2. Creating a Pull Request (PR):
   • Submit PRs to help improve the documentation, add features, or fix issues.

🌐 MCP Ecosystem & Resources

GoMCPTest is part of a larger MCP ecosystem that includes various tools and clients designed for different use cases:

• Visit https://owulveryck.github.io/gomcptest/ to explore additional resources, including experimental documentation.
• For more detailed information on the Model Context Protocol (MCP), refer to official MCP documentation and community forums.

By leveraging GoMCPTest, developers can accelerate the development of advanced AI applications and integrate them into a wide range of workflows with enhanced efficiency and functionality.