mcp -google-adk-sse-multi-tool-system

This MCP server provides a TypeScript/Express-based API toolkit that exposes a set of tools for file operations, making external API calls, managing session data, and retrieving weather information. It is designed to be used in tandem with the Google ADK Agent, a Python client that communicates with the MCP server via a webhook interface. The ADK agent determines which tool to invoke based on user requests and forwards those requests to the MCP server, which executes the operation and returns results for presentation to the user. In practice, you can leverage the server to store and retrieve files, perform HTTP API calls, manage session data across interactions, and fetch weather data as part of automated workflows. The Google ADK agent mediates user intents and translates them into tool invocations that the MCP server can perform, enabling a modular, extensible tool system.

Prerequisites:

• Node.js v16+ for the MCP server
• Python v3.9+ for the Google ADK agent
• Git access tokens if you need to access private repos from the MCP server

Installation steps:

1. Clone the repository: git clone <repository-url> cd <repo-directory>

2. Install dependencies for the MCP server (Node.js): npm install

3. Install dependencies for the Google ADK agent (Python): cd mcp_agent pip install -r requirements.txt

4. Set up environment variables (example):

   • Create a .env file for the MCP server (in mcp/): PORT=9000 BASE_DIR=./data GITHUB_ACCESS_TOKEN=your_github_token_here GITLAB_ACCESS_TOKEN=your_gitlab_token_here REPO_DIR=./repos MAX_EVENT_LISTENERS=100

   • Create a .env file for the Google ADK agent (in mcp_agent/): GOOGLE_CLOUD_PROJECT=your-google-project-id GOOGLE_CLOUD_LOCATION=us-central1 GOOGLE_GENAI_USE_VERTEXAI=True MCP_SERVER_URL=http://localhost:9000

5. Run the MCP server and Google ADK agent: bash ./scripts/run-mcp.sh bash ./scripts/run-agent.sh web

6. Open the web interface to test the system: Visit http://localhost:8000 and select mcp_agent from the dropdown to interact with the agent via the web UI.

Tips and considerations:

• Ensure the MCP server has access to any required private repositories via the GITHUB_ACCESS_TOKEN and GITLAB_ACCESS_TOKEN.
• If you extend the toolset, update the webhook handlers accordingly in the MCP server and expose any new endpoints to the Google ADK agent.
• For Kubernetes deployments, build and push the ADK agent image as described in the README, and configure MCP_SERVER_URL to the appropriate service URL.
• Monitor logs for troubleshooting: look for startup errors related to port bindings, environment variable misconfigurations, or missing tool handlers.
• When adding new tools, ensure input validation and error handling are robust to prevent cascading failures in the ADK workflow.