Fetch.ai has published a runnable A2A Cart Store integration connecting an agent-operated storefront with its uAgents framework and Skyfire’s payment infrastructure. The release demonstrates how an AI agent can move from product selection to completed payment through standardized protocol handoffs, rather than requiring a bespoke checkout integration for every agent ecosystem. Fetch.ai highlighted the example on August 3.
The implementation links the Google-originated Agent2Agent protocol and Google’s Agent Payments Protocol with Fetch.ai’s Payment Protocol through an outbound uAgents adapter. It is a developer reference implementation, not evidence that autonomous purchasing has been activated across the entire Fetch.ai network. The code and setup instructions are publicly available for developers to run and modify.
Agentic commerce cannot stop at product discovery.
Agents also need an interoperable way to complete the transaction.
The A2A Cart Store example connects:
A2A/AP2 store → uAgents adapter → https://t.co/kJ9URVpOul Payment Protocol → @trySkyfire Pay
Under the hood:… pic.twitter.com/gNbCSGqqCL
— Fetch.ai (@Fetch_ai) August 2, 2026
Adapter Translates Commerce and Payment Messages
The store exposes five basic commands: list, add, remove, cart and checkout. When checkout begins, the A2A store creates an AP2 CartMandate containing information such as the total, expiry and payment method. The adapter converts that mandate into a uAgents RequestPayment message, allowing an agent interface using Fetch.ai’s protocols to continue the transaction.
The reverse handoff occurs after the buyer accepts the payment request. A uAgents CommitPayment message becomes an AP2 PaymentMandate, which is returned to the store for validation and charging. A successful charge maps to CompletePayment, while a failed charge maps to CancelPayment; a rejected request denies the cart mandate instead.
Skyfire is the payment method configured in the sample. Its token system lets buyer agents create scoped payment credentials that sellers can verify and charge through an API. The example uses Skyfire to authorize and settle checkout after approval, but Skyfire is not limited to blockchain payments and can support funding through cards, bank transfers or USDC.
Automation Still Depends on Authorization and Controls
The integration reduces the translation work required between agent frameworks, but it does not remove the need for payment authorization. Google describes AP2 as a system of typed, cryptographically verifiable mandates that records who approved a purchase, its limits and its relationship to a specific cart. The objective is controlled agent autonomy rather than unrestricted spending.
A2A handles communication between agents built with different frameworks, while AP2 provides the payment-authorization layer. Fetch.ai’s adapter sits between those standards and the uAgents Payment Protocol, converting structured messages without requiring the underlying agents to share their private logic. That separation could make agent-commerce systems easier to combine while preserving distinct execution and payment components.
The sample is available through Fetch.ai’s Innovation Lab repository and documentation, with environment variables required for the Skyfire seller account, service credentials and charge endpoint. Its immediate value is as testable developer infrastructure, not as proof of commercial adoption, transaction volume or production-scale reliability.
Fetch.ai’s release moves agentic commerce from a conceptual diagram into code that developers can inspect and run. The next test will be whether merchants and agent builders adopt the bridge in live services, where payment failures, refunds, identity checks and authorization policies become more demanding than a controlled cart-store example.