Coinbase Tested AI Agent Stablecoin Payments

A new protocol aims to replace traditional banking fees with machine-to-machine micropayments.

Updated on Oct. 8, 2026 in Artificial Intelligence

Isometric editorial illustration of metallic server hardware blades arranged in a grid, representing autonomous digital infrastructure.
The Coinbase Institute has demonstrated a new protocol using the Base network to enable autonomous AI agents to settle micro-transactions instantly. AI Illustration. Upload story photo >

Live Poll

Do you believe automated machine payments will make using digital services faster and more convenient?

The Coinbase Institute has published a technical paper detailing a method for AI agents to settle micro-transactions using the Base network. Researchers successfully demonstrated a 0.01 USDC transfer that settled in two seconds using the x402 protocol.

Why it matters

The development addresses the economic unviability of legacy banking for AI-driven commerce, where flat transaction fees can represent an overhead of up to 30,000% on small API queries. This architecture enables autonomous agents to interact with web services without needing traditional user accounts or persistent sessions.

The x402 protocol utilizes HTTP 402 Payment Required status codes and EIP-3009 authorization signatures to facilitate automated transfers. In a benchmark on the Base network, researchers recorded a settlement time of 2 seconds for a 0.01 USDC transaction.

The players

Coinbase

A major digital asset exchange providing the infrastructure and research capabilities for the Base network and the x402 protocol.

Citi

A global financial institution that expanded its partnership with Coinbase to integrate traditional banking operations with blockchain systems.

The details

The mechanism functions by intercepting an HTTP 402 status—an error code indicating that a payment is required to access a resource—and automatically signing a transfer authorization via EIP-3009, a standard for gasless token transfers. This process allows software agents to complete payments in a machine-to-machine loop without human intervention or the overhead of maintaining a standard banking session. The research identifies the traditional flat-fee banking model as the primary bottleneck for micropayments, proposing blockchain-based settlement as a high-efficiency alternative.

Timeline

  1. September 28, 2026: Coinbase expanded a strategic partnership with Citi.

  2. October 7, 2026: The Coinbase Institute released the technical paper on AI agent payment protocols.

The Tech Race

This effort follows the Coinbase Institute's recent research into the economic integration of autonomous agents within existing web standards. It marks a departure from traditional financial reliance on flat-fee banking infrastructures, positioning the firm to lead the development of native payment layers for AI-driven platforms.

Current banking systems remain unsuitable for AI agents due to prohibitive fee structures on micro-transactions. This technology aims to eventually enable seamless, automated service subscriptions for software, though broad adoption depends on the outcome of future technical trials.

The takeaway

The transition to machine-based commerce hinges on reducing transaction costs below the micro-penny scale that legacy finance cannot accommodate. Watch for ongoing technical trial results throughout the coming quarters to see if this architecture gains adoption among major web service providers.

Further reading

Learn more about the latest developments in Artificial Intelligence.

Live Poll

Do you believe automated machine payments will make using digital services faster and more convenient?