AI agents settle $73M in 176M crypto transactions

A report from market maker Keyrock, produced with Coinbase and the blockchain Tempo, found AI agents settled $73 million across 176 million crypto transactions between May 2025 and April 2026. More than 98% of agent settlements used Circle’s USDC.

Keyrock researcher Ben Harvey wrote that machine-to-machine payments moved from concept to a functioning ecosystem over the past year. The dataset covers payments executed by autonomous agents that call APIs, interact with protocols and complete small-value transactions on users’ behalf.

By the end of the first quarter of 2026 the report identified more than 104,000 agents listed across 15 or more registries. The average transaction size was about $0.31.

The report points to the economics behind the shift to stablecoins. Traditional card rails impose a fixed processing fee of roughly 30 cents, which makes sub-dollar payments uneconomical. Harvey gave an example: an agent paying three cents for a weather API call cannot route that payment through Visa.

Keyrock reported that incumbents have spent more than $8 billion on acquisitions aimed at building a new payments stack. Use cases for agents include building decentralized applications, launching tokens, interacting with protocols autonomously and experimenting with trading strategies.

A survey of 2,632 crypto users cited in the report found 87% would allow AI to manage at least 10% of their crypto holdings. The report also referenced industry forecasts that AI agents could scale to billions of instances operating with stablecoins within five years.

The report highlighted concentration risk from the heavy use of USDC, noting dependence on a single issuer’s reserve management, regulatory standing and technical infrastructure. It said regulatory action, a de-peg event or sustained downtime at Circle would leave agent payments without an obvious fallback.

Keyrock recommended attention to resilience and diversification in settlement rails as machine-to-machine payments scale, given the small average ticket sizes and current settlement patterns.