This sample demonstrates the A2A ap2-extension for a human present transaction using a card as the payment method.

Human-Present flows refer to all commerce flows where the user is present to confirm the details of what is being purchased, and what payment method is to be used. The user attesting to the details of the purchase allows all parties to have high confidence of the transaction.

The IntentMandate is still leveraged to share the appropriate information with Merchant Agents. This is to maintain consistency across Human-Present and Human-Not-Present flows.

All Human-Present purchases will have a user-signed PaymentMandate authorizing the purchase.

This sample consists of:

• Shopping Agent: The main orchestrator that handles user's requests to shop and delegates tasks to specialized agents.
• Merchant Agent: An agent that handles product queries from the shopping agent.
• Merchant Payment Processor Agent: An agent that takes payments on behalf of the merchant.
• Credentials Provider Agent: The credentials provider is the holder of a user's payment credentials. As such, it serves two primary roles:
  • It provides the shopping agent the list of payment methods available in a user's wallet.
  • It facilitates payment between the shopping agent and a merchant's payment processor.

1. Card purchase with DPAN

• The merchant agent will advertise support for card purchases through its agent card and through the CartMandate once shopping is complete.
• The preferred payment method in the user's wallet will be a tokenized (DPAN) card.

2. OTP Challenge

• The merchant payment processor agent will request an OTP challenge of the user in order to complete payment.

Ensure you have obtained a Google API key from Google AI Studio. Then declare the GOOGLE_API_KEY variable in one of two ways.

• Option 1: Declare it as an environment variable: export GOOGLE_API_KEY=your_key
• Option 2: Put it into an .env file at the root of your repository. echo "GOOGLE_API_KEY=your_key" > .env

You can execute the following command to run all of the steps in one terminal:

bash code/samples/python/scenarios/a2a/human-present/cards/run.sh

Or you can run each server in its own terminal:

1. Start the Merchant Agent:

   uv run --package ap2-samples python -m roles.merchant_agent

2. Start the Credentials Provider:

   uv run --package ap2-samples python -m roles.credentials_provider_agent

3. Start the Merchant Payment Processor Agent:

   uv run --package ap2-samples python -m roles.merchant_payment_processor_agent

4. Start the Shopping Agent:

   uv run --package ap2-samples adk web code/samples/python/src/roles

Open a browser and navigate to the shopping agent UI at http://0.0.0.0:8000. You may now begin interacting with the Shopping Agent.

This section walks you through a typical interaction with the sample.

1. Launching Agent Development Kit UI: Open a browser on your computer and navigate to 0.0.0.0:8000/dev-ui. Select shopping_agent from the Select an agent drop down in the upper left hand corner.
2. Initial Request: In the Shopping Agent's terminal, you'll be prompted to start a conversation. You can type something like: "I want to buy a coffee maker."
3. Product Search: The Shopping Agent will delegate to the Merchant Agent, which will find products matching your intent and present you with options contained in CartMandates.
4. Cart Creation: The Merchant Agent will create one or more CartMandates and share it with the Shopping Agent. Each CartMandate is signed by the Merchant, ensuring the offer to the user is accurate.
5. Product Selection The Shopping Agent will present the user with the set of products to choose from.
6. Link Credential Provider: The Shopping Agent will prompt you to link your preferred Credential Provider in order to access you available payment methods.
7. Payment Method Selection: After you select a cart, the Shopping Agent will show you a list of available payment methods from the Credentials Provider Agent. You will select a payment method.
8. PaymentMandate creation: The Shopping Agent will package the cart and transaction information in a PaymentMandate and ask you to sign the mandate. It will initiate payment using the PaymentMandate.
9. OTP Challenge: The Merchant Payment Processor will then request an OTP, and you'll be asked to provide a mock OTP to the agent. Use 123
10. Purchase Complete: Once the OTP is provided, the payment will be processed, and you'll receive a confirmation message and a digital receipt.

If you want to understand what the agents are doing internally or inspect the mandate objects they create and share, you can ask the Shopping Agent to run in verbose mode.

Enabling verbose mode will instruct the Shopping Agent, and any agents it delegates to, to provide detailed explanations of their process, including:

• A description of their current and next steps.
• The JSON representation of all data payloads (such as IntentMandates, CartMandates, or PaymentMandates) being created, sent, or received.

To activate this mode, simply include the keyword verbose in your initial prompt to the Shopping Agent. Example prompt:

"I'm looking to buy a new pair of shoes. Could you be verbose as we do this, explaining what you're doing, and display all data payloads?"

💡 TIP: Give elaborate instructions

While the word verbose is usually sufficient, providing more elaborate instruction in your prompt tends to result in more detailed and helpful explanations from the agent.

💡 TIP: If the JSON is missing...

If the agent is in verbose mode but fails to display the JSON mandate, a quick follow-up prompt is often needed. Just say: "Remember we're in verbose mode, please display the JSON." After this reminder, the agent usually becomes more reliable at displaying all data payloads.

To help engineers visualize the exact communication occurring between the agent servers, a detailed log file is created automatically when the servers start up.

By default, this log file is named watch.log and is located in the .logs directory.

The watch log is a comprehensive trace that includes three main categories of data: