On a live stage at the India AI Impact Summit, Mastercard turned a checkout counter into a real‑time demo, completing India’s first fully authenticated agentic commerce transaction. The payment ran through Mastercard’s global network, was triggered inside a large language model, and used the open‑source Context Model Protocol to keep card data hidden. This shows how AI can handle checkout without a click.
How the Demo Works
Live Checkout Inside an LLM
The demonstration used Mastercard cards issued by Axis Bank and RBL Bank to purchase items from merchants such as Swiggy Instamart, Vodafone Idea, Tira, and Zepto. When a shopper voiced a request, the LLM translated the intent into a payment request, which then flowed through payment aggregators Cashfree Payments, Juspay, PayU, and Razorpay. You can see how the entire flow stayed within a single conversational thread.
Tokenisation and the Context Model Protocol
Instead of exposing card numbers, the system relied on tokenisation—unique digital identifiers replace sensitive details. The Context Model Protocol (CMP) acted as a secure bridge, allowing the AI to request a tokenised payment while proving its authenticity with cryptographic proofs. This design keeps fraud risk low and data privacy high.
Why It Matters for Payments
Speeding Up the Buyer Journey
AI‑driven checkout can shave seconds off the purchase path, reducing cart abandonment and keeping shoppers engaged. When the payment happens behind the scenes, users stay in the conversation and don’t need to switch apps or enter details manually.
Security Benefits
Tokenisation already cuts card‑present fraud, and CMP adds an extra layer by ensuring that only verified AI agents can trigger payments. The approach also limits exposure of transaction metadata, a key concern for regulators and consumers alike.
Industry Perspective
Technical Insights from a Security Architect
Rohit Mehta, a senior security architect, praised the elegance of linking an LLM with payment rails via CMP. He noted that the protocol’s ability to work across multiple aggregators proves its flexibility, and he stressed that robust audit and monitoring tools will be essential as AI‑initiated calls move into production.
Impact on Merchants and Consumers
What Merchants Can Expect
Retailers like Instamart could let you order groceries by simply stating a need, while the AI handles payment instantly. This end‑to‑end flow promises smoother conversions and lower operational overhead for merchants.
What Consumers Need to Know
While the convenience is appealing, you’ll want assurance that your privacy is protected. Mastercard’s reliance on tokenisation and open‑source authentication protocols aims to build that trust, but widespread awareness of these mechanisms will be crucial for adoption.
Next Steps for Mastercard
Mastercard plans to roll the Agentic Commerce Framework out across the Asia‑Pacific region, piloting with additional merchants and expanding the list of participating banks. The real challenge now is moving from proof‑of‑concept to a secure, scalable reality that regulators approve and that you can rely on for everyday purchases.
