Mastercard Just Validated the Standard We Built: Verifiable Agent Actions with AAR

On March 5, 2026, Mastercard announced Verifiable Intent : an open-source cryptographic framework for proving AI agent transactions were authorized and executed as intended. Google, IBM, Fiserv, and Checkout.com signed on immediately. That announcement matters because it confirms a hard truth many of us already hit in production: agent payments and autonomous actions cannot scale on trust alone. If an agent can move money, place trades, call APIs, and trigger external workflows, you need cryptographic evidence of what actually happened. We built that independently before this announcement and shipped it live: Agent Action Receipts (AAR) . It's an open spec, MIT-licensed SDK, and it's already attached to real API responses in production. This post covers the problem, the standard, and the exact TypeScript code to implement and verify receipts. The Problem: Agents Act, But Nobody Can Prove It AI agents now execute real-world operations: transfer stablecoins, rebalance portfolios, submit