Visa and Artemis Analytics released a report titled 'Agentic Payments from the Ground Up,' analyzing the protocols and market structure forming around AI-initiated payments. The report identifies the catalyst as a demand threshold crossed by AI capability since mid-2025 frontier model releases, enabling agents to autonomously discover APIs, evaluate pricing, and execute payments without human intervention. Two open-source protocols have emerged to address infrastructure gaps: the x402 protocol and the Machine Payments Protocol (MPP), both designed for high-frequency, sub-dollar transactions between software systems.
The report distinguishes two transaction categories: macro-level payments, where agents act as proxies for human intent in purchases such as travel bookings or subscription management; and micro-level payments, characterized by high-frequency, sub-dollar exchanges between software systems including API calls, data queries, and compute access, often valued at less than one cent per transaction. The authors argue that the latter category requires fundamentally new payment infrastructure.
x402 and MPP Protocols Process Millions in AI-Driven Transactions
The x402 protocol, originally developed by Coinbase and Cloudflare and now governed by the Linux Foundation, processed over $15 million in adjusted volume across 109.6 million adjusted transactions since its May 2025 launch. Base blockchain accounted for approximately 90% of activity. The Machine Payments Protocol (MPP), developed by Stripe and Tempo with Visa's involvement, recorded roughly 115,000 adjusted transactions and $25,000 in adjusted volume since its March 2026 launch, averaging around 4,000 transactions per day.
Trust and Compliance Gaps Identified as Primary Challenges
The report identifies agent trust as the most pressing unsolved problem in agentic commerce. Software agents can execute hundreds of erroneous or compromised transactions before any oversight mechanism detects a failure. Four distinct risk categories exist: mis-purchase errors, adversarial prompt injection attacks, unclear liability chains, and cascading failures in multi-agent transaction networks. Existing legal and regulatory frameworks were not designed for delegated machine-to-machine commerce, and no jurisdiction has established clear precedent for distributing responsibility across the human principal, agent platform, model provider, and merchant.
The study frames these gaps as the primary opportunity for established payments institutions. Card networks, issuers, and acquirers possess decades of accumulated infrastructure in identity verification, fraud detection, cross-border settlement, and dispute resolution — capabilities that crypto-native protocols currently lack at scale. The report concludes that the institutions best positioned to define the next generation of agentic payment infrastructure are those that can combine protocol-native speed with compliance-grade trust, merchant distribution, and interoperability across both stablecoin and card-based settlement rails.
FAQ
What did Visa and Artemis Analytics release?
Visa and Artemis Analytics released a report titled 'Agentic Payments from the Ground Up,' providing the first comprehensive data-driven analysis of the protocols, standards, and market structure forming around AI-initiated payments.
How much volume has the x402 protocol processed since launch?
The x402 protocol processed over $15 million in adjusted volume across 109.6 million adjusted transactions since its May 2025 launch, with Base blockchain accounting for approximately 90% of activity.
What are the four risk categories identified in agentic commerce?
The report identifies four distinct risk categories: mis-purchase errors, adversarial prompt injection attacks, unclear liability chains, and cascading failures in multi-agent transaction networks.
