ConsultingAgents

You can run a dedicated MCP server that lets Claude Code consult with multiple expert agents for code analysis and problem solving. This server exposes four consultants and supports both local (stdio) and HTTP transport, enabling you to get diverse perspectives and richer results when debugging or designing code.

How to use

Use the MCP client in Claude Code to register and connect to the ConsultingAgents server. Once connected, you can invoke consult_with_darren, consult_with_sonny, consult_with_sergey, and consult_with_gemma to get expert input on coding tasks. You can supply a clear consultation context and, when helpful, include the relevant source code or repository URL to enrich the analysis. The server will coordinate the responses from Darren, Sonny, Sergey, and Gemma and return a consolidated set of recommendations.

How to install

Prerequisites: you need Python 3.8 or newer and access keys for the respective AI services.

Create a working directory and set up a Python virtual environment to isolate dependencies.

Install the required Python packages from the project requirements.

Prepare API keys by placing them in a .env file at the project root.

Start the MCP server using the provided script.

# Clone and enter the project dir
# (Replace with your actual clone path if different)
git clone https://github.com/matthewpdingle/consulting-agents-mcp.git
cd consulting-agents-mcp

# Set up a Python virtual environment
python -m venv mcp_venv
source mcp_venv/bin/activate  # On Windows: mcp_venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Create environment file with API keys
cat > .env << 'EOF'
OPENAI_API_KEY=your_openai_api_key_here
ANTHROPIC_API_KEY=your_anthropic_api_key_here
GOOGLE_API_KEY=your_google_api_key_here
EOF

# Start the MCP server
chmod +x start_mcp_server.sh
./start_mcp_server.sh

Configuration and security

The server uses environment variables for API keys and transport settings. Ensure your API keys are kept secure and not committed to version control.

Environment variables you will typically set include the API keys shown above as well as transport configuration if you choose the HTTP or SSE transport.

Troubleshooting

If the MCP server cannot be found by Claude Code, verify that you registered the server with the correct absolute path to the start script.

If you encounter authentication errors, double-check that your API keys are correctly set in the .env file and that the server is running.

If you experience connection issues, ensure the MCP server is active and listening on the expected transport (stdio by default) and that Claude Code is configured to connect to it.

Development notes

For development, you can run with debug output to observe internal processing and tool responses. Test the consult endpoints locally if you enable HTTP transport.

Project structure

Key files include the MCP server implementation, the startup script, and dependency definitions. These provide the runtime for coordinating the four consulting agents.

Notes on usage with Claude Code

After starting the server, register it with Claude Code and enable the MCP integration to access the four consulting tools from within your coding workflow.

Available tools

consult_with_darren

Uses OpenAI's o3-mini model to analyze code and provide recommendations.

consult_with_sonny

Uses Claude 3.7 Sonnet with enhanced thinking to provide in-depth code analysis.

consult_with_sergey

Uses GPT-4o with web search to find relevant documentation and examples.

consult_with_gemma

Uses Google's Gemini 2.5 Pro to analyze repositories and craft development plans.