India is preparing a framework to let artificial intelligence agents make small digital payments on the Unified Payments Interface without approval for each transaction. The framework may be unveiled by the National Payments Corporation of India as the Unified Agent Protocol at the 2026 edition of the Global Fintech Fest in Mumbai. The move could place UPI at the center of agentic commerce.
The Scale of UPI
UPI processed 24.51 billion transactions worth 29.8 trillion rupees, or about 314.21 billion US dollars, in August. A 2025 International Monetary Fund report called it the world's largest retail fast-payment system by transaction volume. This scale makes UPI a natural platform for testing how AI agents could participate in everyday digital commerce.
A National Framework for Agentic Payments
According to a Reuters report citing three sources, the proposed system would let customers define rule-based instructions for AI agents covering when and how much to pay. These instructions would include spending limits, identity checks and audit trails designed to keep delegated authority within safe boundaries. The sources added that the National Payments Corporation of India also plans to introduce a liability framework, though no details have been disclosed.
How Delegated Payments Would Work
Rather than requiring manual authorization for each transaction, the framework would permit an AI agent to act within pre-set limits. A user could instruct an agent to purchase groceries within a fixed budget, and the agent would complete the UPI payment once conditions are met. The design aims to shift UPI from a system where people directly approve every payment to one where limited authority can be delegated to software.
Building on Existing UPI Mechanisms
The protocol is expected to leverage UPI Circle and Reserve Pay, two existing mechanisms that already support delegated and blocked funds. UPI Circle allows a primary user to authorize trusted individuals to make payments on their behalf, while Reserve Pay lets customers block funds for multiple future debits. According to the Reuters report, the amount and validity period for such blocks could be reconsidered to support agentic payments.
Early Use Cases and Merchant Integration
Initial applications are likely to focus on low-value and high-frequency transactions such as grocery orders and routine digital purchases. Over time, sources said NPCI expects AI agents to handle more sophisticated use cases like placing orders automatically based on sales or discounts and potentially making investments at specified price thresholds. The payments body also plans to provide merchants with infrastructure to integrate directly with the protocol.
India's Agentic Payments Race
The move would place UPI alongside efforts by Visa and Mastercard, which are building payment capabilities for AI agents in India. Mastercard said it completed its first authenticated agentic transaction in New Delhi in June. Domestic fintech companies are also active, with Pine Labs launching its P3P protocol and Razorpay and Cashfree exploring agentic commerce and in-chat transactions.
Safeguards and Liability Considerations
Allowing AI agents to move money raises new questions about security, errors and unauthorized transactions. The expected safeguards include spending limits, identity checks and audit trails, while NPCI plans to build a liability framework for cases where an agent pays incorrectly or exceeds its instructions. Details on how liability would be assigned have not yet been made public.
The proposed framework signals a broader shift from AI as a recommendation tool to AI as a transaction participant within India's digital payments ecosystem. If implemented, it would give consumers the ability to delegate routine financial tasks while maintaining control through predefined limits and audit mechanisms. More details are expected when NPCI presents the Unified Agent Protocol at the Global Fintech Fest.