According to Woofun AI, AI agent payments are being hailed as the disruptive financial innovation following contactless payments, but a BeInCrypto investigation into the x402 protocol reveals the true picture of how it operates at its foundation. In its early stage of adoption, this heavily hyped technology displays micro-level characteristics entirely at odds with the grand narrative.
Between late July and August 2026, the x402 protocol recorded 6.4 million payments, with every individual transaction below US$120,000. Data compiled by Woofun AI shows that 90% of these payments were worth less than 1 US cent, highlighting an extreme micro-transaction trait. Among specific use cases, AI-generated cryptocurrency security assessment services, priced at 2 cents per query, became the most frequent application; web search services followed closely behind, with AI agents paying per use to obtain the information needed for their tasks; flight query services were also common, with assistants planning vacations first paying a small fee to check flights, then paying again to retrieve destination information, thereby constructing a complete decision-making chain.
On the infrastructure side, the Solana network demonstrated absolute dominance, carrying 99.2% of transaction traffic. This concentration is even more striking when viewed over time: within just 8 days, transactions on Solana accounted for 94.5% of all recorded payments. It is worth noting that the high transaction volume may stem from a small number of active wallets rather than broad enterprise adoption.
In the suitability assessment, micro-payments ranked first with a score of 60.7, while trading and investment AI agents scored only 39.1, and institutional-grade workflows scored as low as 12. Among the 12 institutional-grade processes evaluated, only 6 were able to complete payments within preset permissions, and concrete evidence of real-world application remains scarce.
Funding limits are directly tied to risk exposure. Will Papper, Director of AI Agent Payment Products at Cloudflare (NET.US), pointed out that if an AI manages only US$10, the spending risk is controllable; but if the limit rises to US$1,000, extreme caution is required. This view underscores the necessity of permission management. Before granting AI agents spending authority, enterprises should conduct in-depth assessments by referring to the full report.