Mastercard's Live Agentic Payments Launch in Singapore
Mastercard launches live agentic payments in Singapore with DBS and UOB, enabling AI agents to book services and process payments autonomously using Agent Pay framework.
The first live authenticated agent-based payment transaction has gone live in Singapore through Mastercard, DBS, and UOB. This marks a critical shift from proof-of-concept demonstrations to production-ready autonomous commerce infrastructure.
The transaction involved an AI agent booking airport transportation through hoppa's mobility network, with payment processing handled entirely autonomously. No human confirmation clicks, no manual payment steps — just agent-initiated commerce from start to finish.
Agent Pay Architecture
The underlying infrastructure runs on Mastercard Agent Pay, a framework designed specifically for AI-initiated purchases. The system implements three core security layers that address the fundamental trust problem in autonomous payments.
Each transaction uses a dedicated Mastercard Agentic Token issued per agent, establishing a clear identity and permission boundary. Consumer consent gets captured explicitly upfront rather than per-transaction, while purchase confirmation flows through Mastercard Payment Passkeys for cryptographic authentication.
- Agentic Tokens — unique identifiers per AI agent with scoped permissions
- Payment Passkeys — cryptographic authentication replacing manual confirmations
- Explicit Consent — upfront authorization for agent spending within defined parameters
- Tokenized Credentials — secure storage of payment data with agent-specific access controls
Technical Implementation Details
CardInfoLink's AI agent handled the booking logic, connecting to hoppa's transportation network APIs. The agent evaluated available options, selected appropriate services, and executed payment — all within preset spending limits and service categories.
The authentication flow combines traditional payment security with agent-specific controls:
- Identity verification — passkey-based consumer authentication
- Agent authorization — token-based permission validation
- Transaction limits — predefined spending boundaries per agent
- Service scope — category restrictions on what agents can purchase
Regional Deployment Strategy
Mastercard is establishing a dedicated AI Centre of Excellence in Singapore, with specialized agentic commerce teams across APAC. The company has already completed similar authenticated transactions in Australia, New Zealand, and India, but Singapore represents a key strategic hub.
Notably, DBS participated in a separate agentic payments pilot with Visa in February 2026, executing food and beverage transactions using AI agents. The same bank appearing in competing payment networks' agent initiatives within weeks signals aggressive institutional positioning.
Production Use Cases
The initial transportation booking demonstrates a clear pattern for autonomous commerce expansion. Mastercard plans to extend Agent Pay across sectors where manual payment friction creates user experience bottlenecks.
Target verticals include:
- Transportation — ride booking, parking payments, transit passes
- Travel — hotel reservations, flight changes, travel insurance
- Retail — subscription renewals, recurring purchases, inventory restocking
Security and Trust Considerations
The authentication architecture addresses the core challenge of autonomous payments: how to maintain security and consumer control when removing human confirmation steps. Agent Pay shifts the consent model from per-transaction to per-agent authorization.
This creates new attack vectors and trust requirements. Agents need robust spending controls, clear audit trails, and fail-safe mechanisms when services change or become unavailable. The passkey implementation provides cryptographic security, but operational security depends on proper agent configuration and monitoring.
Infrastructure Implications
Live agentic payments require more than authentication — they need reliable agent-to-service communication, standardized API interfaces, and consistent error handling. The Singapore implementation suggests these infrastructure pieces are reaching production readiness.
Payment networks are positioning themselves as the trust layer for autonomous commerce, similar to how they function in traditional e-commerce. The competition between Mastercard and Visa for agentic payment dominance is playing out through bank partnerships and pilot programs.
Bottom Line
This isn't just another payments demo — it's production infrastructure for autonomous commerce. The fact that major Asian banks are running multiple agentic payment pilots simultaneously indicates serious institutional commitment to agent-driven transactions.
The technical architecture solves real problems around authentication and consent in autonomous systems. For developers building AI agents that need to transact, Agent Pay provides a clear path from prototype to production without rebuilding payment infrastructure.