How to Automate Google Pay Integrations With MCP

Google recently released the Google Pay & Wallet Developer MCP server in Public Preview to connect AI assistants directly with real-time merchant account context. Built on the Model Context Protocol (MCP) specification version 2025-11-25, this server allows your IDE to search official documentation, validate integration code, and monitor payment performance without requiring you to switch tabs.

This guide covers how to set up the server, configure the necessary identity access, and expose its functionality to AI environments like VS Code, Google AI Studio, or Cursor.

Architecture and Security Constraints

The server operates as a remote Google Cloud service rather than a local binary. It requires a dedicated Google Cloud project with the Pay & Wallet Developer API enabled via the paydeveloper.googleapis.com endpoint.

Unlike many early MCP implementations, this server does not accept static API keys. Authentication relies entirely on OAuth 2.0 and Google Cloud Identity and Access Management (IAM). You must provision a distinct IAM identity specifically for your AI agent. This ensures that all actions taken by the model are logged and scoped to the exact permissions required for integration management.

Data flow between your IDE and the Google Pay APIs is governed by Model Armor, a Google Cloud security layer. Model Armor intercepts and scans all incoming MCP calls and outbound responses to prevent data leakage or the execution of malicious prompts.

Setting Up the Environment

To begin using the server, you need to configure your Google Cloud infrastructure to accept incoming protocol requests. Start by navigating to your Google Cloud Console and enabling the paydeveloper.googleapis.com service.

Next, configure the OAuth 2.0 consent screen and generate credentials for a desktop application. These credentials will allow your IDE to request an access token on behalf of the developer identity.

The official Google documentation provides the exact scopes required for the OAuth flow. Your agent will need permissions that cover both read operations for performance metrics and write operations for enabling merchant APIs.

Core Tool Capabilities

Once connected, the server exposes four primary functional areas to your AI assistant. These tools allow the model to autonomously retrieve information and execute management tasks.

Functional Area	Description
Documentation Search	The search_documentation tool retrieves the latest code samples, best practices, and integration guidance directly from Google’s official repositories.
Integration Management	Models can automatically enable APIs for specific merchant accounts and query the status of existing integration endpoints.
Performance Monitoring	Exposes aggregated performance data and error metrics for merchant accounts, supporting a lookback window of up to 30 days.
Passes Management	Lists Google Wallet pass issuers and classes such as loyalty cards and boarding passes, alongside tools to validate pass definitions.

When a model needs to troubleshoot a failed payment integration, it can use the performance monitoring tools to retrieve the last 30 days of error metrics. If the error relates to an outdated API call, the model can invoke search_documentation to find the correct implementation and immediately suggest the code change.

Agentic Commerce Integration

This server is a component of Google’s broader “agentic commerce” ecosystem introduced at Google I/O 2026. This architecture shifts digital commerce from human-driven UI interactions to automated, agent-led transactions.

If you are building consumer-facing AI agents, you will use this MCP server alongside the Universal Commerce Protocol (UCP). UCP standardizes how agents communicate with payment gateways to handle checkout flows automatically.

To support real-time adjustments during agent-led transactions, the Android Pay API now includes new dynamic callbacks. The onPaymentAuthorized and onPaymentDataChanged callbacks allow your application to adjust tax calculations or shipping costs instantly while the agent is processing the order.

Security for these autonomous transactions relies on a human-in-the-loop model. When an AI agent running on a desktop environment attempts to authorize a high-value purchase, the system triggers Cross-Device Biometric Authentication. The agent pauses the transaction and requests human verification via a mobile biometric prompt, such as a fingerprint or facial scan on a paired device. Only after receiving the mobile hardware attestation will the payment proceed.

Next Steps

Review your current Google Cloud IAM structure and create a dedicated service account or OAuth identity for your local AI assistant. Map this identity strictly to the required scopes for the paydeveloper.googleapis.com endpoint before connecting your IDE.