Integrate Google's Vertex AI Agent with Live AlloyDB Data via CData Connect AI

Vertex AI provides a development ecosystem for building AI agents using the Agent Development Kit (ADK). ADK enables developers to create tool-augmented agents that can reason, take actions, and interact with external systems through structured tool interfaces. These agents can be tested locally in the ADK Web interface and extended with advanced logic for enterprise workflows.

By integrating Vertex AI ADK with CData Connect AI through the built-in MCP (Model Context Protocol) Server, your agents gain the ability to query, analyze, and act on live AlloyDB data in real time. This connection bridges Google's agent-building framework with the governed enterprise connectivity of CData Connect AI, ensuring every request runs securely against authorized data sources without manual data movement.

This article outlines the steps to configure AlloyDB connectivity in Connect AI, generate the required authentication token, configure the Vertex AI ADK environment, and verify that your agent can successfully communicate with live AlloyDB data through the CData MCP Server.

Step 1: Configure AlloyDB connectivity for Vertex AI

Connectivity to AlloyDB from Vertex AI is made possible through CData Connect AI's Remote MCP Server. To interact with AlloyDB data from Vertex AI, start by creating and configuring a AlloyDB connection in CData Connect AI.

1. Log into Connect AI, click Sources, and then click Add Connection

2. Select AlloyDB from the Add Connection panel

3. Enter the necessary authentication properties to connect to AlloyDB.

   The following connection properties are usually required in order to connect to AlloyDB.

   • Server: The host name or IP of the server hosting the AlloyDB database.
   • User: The user which will be used to authenticate with the AlloyDB server.
   • Password: The password which will be used to authenticate with the AlloyDB server.

   You can also optionally set the following:

   • Database: The database to connect to when connecting to the AlloyDB Server. If this is not set, the user's default database will be used.
   • Port: The port of the server hosting the AlloyDB database. This property is set to 5432 by default.

   Authenticating with Standard Authentication

   Standard authentication (using the user/password combination supplied earlier) is the default form of authentication.

   No further action is required to leverage Standard Authentication to connect.

   Authenticating with pg_hba.conf Auth Schemes

   There are additional methods of authentication available which must be enabled in the pg_hba.conf file on the AlloyDB server.

   Find instructions about authentication setup on the AlloyDB Server here.

   Authenticating with MD5 Authentication

   This authentication method must be enabled by setting the auth-method in the pg_hba.conf file to md5.

   Authenticating with SASL Authentication

   This authentication method must be enabled by setting the auth-method in the pg_hba.conf file to scram-sha-256.

   Authenticating with Kerberos

   The authentication with Kerberos is initiated by AlloyDB Server when the ∏ is trying to connect to it. You should set up Kerberos on the AlloyDB Server to activate this authentication method. Once you have Kerberos authentication set up on the AlloyDB Server, see the Kerberos section of the help documentation for details on how to authenticate with Kerberos.

4. Click Save & Test
5. Navigate to the Permissions tab and update user-based permissions

Add a Personal Access Token

A Personal Access Token (PAT) is used to authenticate the connection to Connect AI from Vertex AI. It is best practice to create a separate PAT for each integration to maintain granular access control.

1. Click the gear icon () at the top right of the Connect AI app to open Settings
2. On the Settings page, go to the Access Tokens section and click Create PAT
3. Give the PAT a descriptive name and click Create

4. Copy the token when displayed and store it securely. It will not be shown again

With the AlloyDB connection configured and a PAT generated, Vertex AI can now connect to AlloyDB data through the CData MCP Server.

Step 2: Install required dependencies

Enable the necessary Google Cloud APIs so Vertex AI ADK can run Gemini models, build agent environments, and access supporting services inside your Google Cloud project. These APIs provide the backend capabilities that ADK relies on during development and execution.

1. Visit the Google Cloud Console
2. Click the Project Picker at the top of the page and choose New Project

3. Create the project and note the Project ID. Save this ID later for environment configuration
4. From the left navigation menu, open APIs & Services and then choose Enabled APIs & Services
5. Click Enable Apis and services

6. Enable the following APIs:
   • Vertex AI API
   • Cloud Build API
   • Artifact Registry API
   • Service Networking API
   • Cloud Logging API

With these services enabled, your Google Cloud project is prepared for Vertex AI ADK development and local tool execution.

Prepare the Vertex AI ADK project folder

Create the project directory and set up the Python environment. This step prepares a clean workspace where ADK installs correctly and loads your agent without dependency conflicts.

1. Open Google Cloud console and select the Cloud Shell. Make sure to select the project you created from the project picker.

2. Create the ADK project directories:

				mkdir -p ~/adk_agents/cdata_mcp_agent
				cd ~/adk_agents/cdata_mcp_agent 
			

3. Create and activate a Python virtual environment:

		python3 -m venv .venv
		source .venv/bin/activate
	

4. Install the required ADK and MCP packages:

				python -m pip install --upgrade pip
				python -m pip install google-adk
				python -m pip install mcp
				python -m pip install --upgrade "google-cloud-aiplatform[agent_engines]"
			

Step 3: Create the ADK Agent files

Define the agent modules so ADK recognizes your agent. This structure allows Vertex AI to load your MCP configuration and register the tools exposed by the CData MCP Server.

1. Create agent.py file and paste the code below in it

		
import os
import base64
import logging
 
from google.adk.agents import LlmAgent
from google.adk.tools.mcp_tool.mcp_toolset import MCPToolset
from google.adk.tools.mcp_tool.mcp_session_manager import StreamableHTTPConnectionParams
 
# ---------- Logging ----------
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
 
# ---------- CData MCP config ----------
CDATA_MCP_URL = os.environ.get("CDATA_MCP_URL", "https://mcp.cloud.cdata.com/mcp")
CDATA_USER_ID = os.environ.get("CDATA_USER_ID")
CDATA_PAT = os.environ.get("CDATA_PAT")
 
tools = []
 
if not (CDATA_USER_ID and CDATA_PAT):
    logger.warning(
        "CData MCP credentials not set (CDATA_USER_ID or CDATA_PAT missing); "
        "starting agent WITHOUT MCP tools."
    )
else:
    # Basic auth header: base64("user:pat")
    basic_auth_bytes = f"{CDATA_USER_ID}:{CDATA_PAT}".encode("utf-8")
    basic_auth_header = base64.b64encode(basic_auth_bytes).decode("utf-8")
 
    try:
        logger.info("Initializing CData MCPToolset against %s", CDATA_MCP_URL)
 
        tools.append(
            MCPToolset(
                connection_params=StreamableHTTPConnectionParams(
                    url=CDATA_MCP_URL,
                    headers={
                        "Authorization": f"Basic {basic_auth_header}",
                        # ADK handles content-type etc. internally;
                        # we just pass auth headers.
                    },
                ),
            )
        )
 
        logger.info("CData MCPToolset initialized successfully.")
    except Exception as e:
        logger.exception("Failed to initialize CData MCPToolset")
 
# ---------- Root agent ----------
root_agent = LlmAgent(
    model="gemini-2.0-flash",
    name="cdata_mcp_agent",
    instruction=(
        "You are a data assistant. Use the CData MCP tools (if available) to "
        "list connections, list catalogs/schemas/tables, and run SQL-style queries."
    ),
    tools=tools,
)
	

2. Create another file __int__.py and paste the code below in it

		
		from .agent import root_agent
		__all__ = ["root_agent"]

Export environment variables

Export the required environment variables to authenticate to Connect AI. These values enable the agent to initialize the MCP toolset and communicate with the CData MCP Server. Before you do that obtain a Google API key so ADK can authenticate to Gemini models. This key enables the agent to run LLM reasoning and route tool calls correctly inside the Vertex AI ADK environment.

1. Visit the Google AI Studio API Key page
2. Click Create API Key. Provide a name for the api-key and choose or create a project if you do not have one

3. Click on Create a key and then copy the API key
4. Return to the Google Cloud Shell page and run the environment variable exports. Replace "your_cdata_email", "your_pat", "your-project-id", and "your_google_api_key" with your values

		export CDATA_MCP_URL="https://mcp.cloud.cdata.com/mcp"
		export CDATA_USER_ID="your_cdata_email"
		export CDATA_PAT="your_pat"
		export GOOGLE_API_KEY="your_google_api_key"
		export VERTEXAI_PROJECT="your-project-id"
		export VERTEXAI_LOCATION="us-central1"
	

Step 4: Launch the ADK web interface

Start the ADK Web interface to load your agent. The interface initializes the runtime and makes your MCP-enabled agent available for interactive testing.

1. Move to the parent folder:

				cd ~/adk_agents
			

2. Launch ADK Web:

				adk web .
			

Step 5: Select your agent and test MCP connectivity

Select your agent from the ADK Web interface. ADK loads the MCP tools and prepares the environment so you issue live MCP queries.

1. Open the ADK Web UI from the browser tab
2. Select cdata_mcp_agent from the agent dropdown

3. Enter list catalogs in the chat panel. ADK returns a live list of your Connect AI connections

At this point, your Vertex AI ADK agent communicates with the CData Connect AI MCP Server and retrieves live AlloyDB data metadata through remote MCP tools.

Get CData Connect AI

To access 300+ SaaS, Big Data, and NoSQL sources directly from your cloud applications, try CData Connect AI today!