This MCP server powers the A2A-ADK security pipeline by exposing MCP-enabled tools for SQL querying, data processing, and privacy masking within a multi-agent framework. The pipeline features three specialized agents: Judge (security threat detection and input sanitization), SQL (database interaction and MCP-driven query tooling), and Mask (PII masking and data privacy transformations). Interactions flow from an A2A client through the Judge, then the SQL agent, and finally the Mask agent before returning results to the client. You can connect to the MCP server from the included client scripts to issue natural-language requests that are translated into MCP tool calls, enabling secure, context-aware data processing while protecting sensitive information.

To use, start the MCP server alongside the ADK and A2A components, then run the client script to send requests. The MCP integration leverages MCP tools to perform SQL-related tasks and data processing within the advisor pipeline. The system is designed to block malicious inputs, analyze queries with Model Context Protocol tooling, and mask PII in the final results.

Prerequisites:

• Python 3.8+ installed on the host
• Poetry or pip (for dependency management) or simply use pip with requirements.txt
• Access to Google ADK and related packages as described in the project documentation

Installation steps:

1. Clone the repository: git clone https://github.com/your-org/rubenszimbres-multi-agent-system-a2a-adk-mcp.git cd rubenszimbres-multi-agent-system-a2a-adk-mcp

2. Create and activate a Python environment (optional but recommended): python3 -m venv venv source venv/bin/activate

3. Install dependencies: pip install -r requirements.txt

   If using a dependency manager like Poetry, adapt accordingly

4. Configure environment variables (see below) as needed for your environment.

5. Start the MCP server to expose the MCP tools for the ADK/A2A integration: python server_mcp.py

6. (Optional) Run the ADK servers and A2A clients as described in the repository's usage docs to exercise the full pipeline.

Environment variables and configuration:

• GOOGLE_API_KEY, GOOGLE_CLOUD_PROJECT, GOOGLE_CLOUD_LOCATION, and GOOGLE_GENAI_USE_VERTEXAI are used when deploying and running ADK/MCP components in Google Cloud environments; set these in your environment or via a secret manager.
• AGENT_PATH and APP_NAME are used by deployment scripts to locate and initialize ADK agents.
• AGENT_PATH should point to the ADK/agents directory containing the agent implementations (e.g., sql_assistant, judge_agent, mask_agent).

Common issues and tips:

• If the MCP server fails to start, verify that server_mcp.py exists at the expected path and that Python can import required MCP tooling modules.
• Ensure port availability if running local servers (the README mentions ports like 10002–10004 for related services in the ADK stack).
• When running in containers, pass the necessary environment variables to the container and ensure volumes (for AGENT_PATH) are mounted correctly.
• If network calls to A2A or MCP services fail, check firewall rules and ensure DNS resolution works for the service endpoints.