NPCI expected to unveil UPI framework for AI-agent payments
NPCI is reportedly set to introduce a Unified Agent Protocol at Mumbai’s Global Fintech Fest, allowing verified AI agents to make preset small UPI payments. The framework could support automated grocery and e-commerce purchases, with delegated-payment limits, verification and dispute safeguards.
What happened
National Payments Corporation of India (NPCI) · NPCI is expected to announce the Unified Agent Protocol at Mumbai’s Global Fintech Fest, enabling verified AI
Key facts
- ₹10,000
- 90 days
- 24.5B monthly transactions
Why this matters
Payments, commerce and AI platforms should evaluate partnerships around agent verification, consent management, transaction monitoring and merchant-side orchestration ahead of the proposed protocol launch.
What to watch
- NPCI publication of the Unified Agent Protocol specifications, certification requirements and initial transaction limits.
- Named bank, PSP, merchant and AI-assistant pilot partners at Global Fintech Fest or subsequent NPCI announcements.
- Whether mandates permit merchant-specific, category-specific or open-loop delegated spending.
- Rules assigning liability among customer, agent provider, bank, PSP, merchant and marketplace for unauthorized or incorrect purchases.
- Launches of agent checkout APIs by major Indian marketplaces, quick-commerce firms, food-delivery platforms and payment aggregators.
- RBI commentary on AI-agent authorization, data sharing, grievance handling, recurring payments and consumer-protection safeguards.
- Evidence that quick-commerce and grocery baskets become the first scaled category, especially replenishment items with predictable preferences.
- Prioritize UPI AutoPay, tokenized mandates and consent-ledger capabilities for recurring baskets and replenishment use cases.
- Create an agent-ready commerce API layer covering SKU availability, dynamic pricing, delivery slots, substitution preferences, cancellation and returns.
- Treat product data quality as a payments issue: normalize pack sizes, inventory accuracy, eligibility rules and promotion terms so agents do not generate disputed orders.
- Design explicit delegated-purchase controls for spend caps, approved categories, merchants, time windows, household members and human-review thresholds.
- Model a shift in customer acquisition from app installs and keyword search toward preferred-agent placement, structured catalog feeds and reliable fulfillment scores.
- Prepare fraud and service workflows for agent-originated disputes, including proof of mandate, agent identity, order instruction history and fulfillment evidence.
Also reported by
- IndianWeb2 — Same time