TRM Labs found that AI agents play a minimal role in online payments on Coinbase's x402 protocol, accounting for only a small fraction of transactions.
TRM Labs Reveals Limited AI Agent Involvement in Payments
TRM Labs, a blockchain intelligence firm, has found that the majority of payment transactions on Coinbase’s x402 protocol are not driven by AI agents. In a report released on Wednesday, the company analyzed approximately $52.7 million across 198.9 million transactions processed via the x402 protocol on Base, Solana, and Polygon since May 2025.
Findings from the Analysis
After filtering out self-payments and irregular transactions, TRM identified around $25.62 million as likely commerce. Despite this figure, they noted that AI agents accounted for only a small fraction of the transactions, between 0.6% and 7.5%. This discrepancy arises from the challenge of distinguishing AI agents from automated scripts, with the latter often mimicking the behavior of agents by executing repetitive transactions.
Coinbase launched the x402 protocol in 2025 to streamline online payments, allowing buyers to receive quotes and authorize payments before a facilitator verifies the transaction. TRM Labs pointed out that typical scripts could perform these tasks without the need for AI agents, indicating that scheduled jobs and self-dealing could produce similar blockchain records.
Refining the Analysis
To enhance the accuracy of their analysis, TRM eliminated self-payments, bulk transactions from a limited number of payers, and sellers with fewer than ten buyers. The resulting commerce figure of $25.62 million was further refined to focus on payments processed by facilitators, with the average transaction amount being less than $1. A more rigorous examination required that payment patterns persisted for several months and involved public agent registration or payments to multiple sellers.
Throughout the observed period, USDC made up 99.6% of the settled value, totaling $52.47 million. Notably, there were instances of meme-token minting and payments to an AI-analysis service in late 2025, followed by a spike in payments for AI services through an agent-payment router in early 2026.
Industry Trends and Challenges
As investments in agent payments increase, notable developments include Binance launching its Agent OS with an x402 payment layer and Coinbase initiating a $1 million accelerator targeting agent and payment startups. Additionally, Amazon introduced AgentCore Payments in collaboration with Coinbase and Stripe, facilitating agent payments for online services using stablecoins.
TRM Labs also discussed the difficulties in attributing responsibility for agent payments, emphasizing the need for better registration processes and reliable counterparty reputation systems. They argued that while the infrastructure exists, improvements in registration and verification are necessary for the growth of agentic commerce.
“Agentic commerce will require agentic compliance,” TRM stated, highlighting the need for systems that manage numerous small payments effectively.
