The proposed standard introduces a 'Job' primitive with escrowed payments, aiming to let AI agents transact onchain without intermediaries.
Imagine a world where AI agents hire other AI agents, pay them, verify the work, and settle up, all without a human touching a single button. That’s the pitch behind ERC-8183, a new Ethereum standard that just got its first official builder session, co-hosted by Virtuals Protocol and the Ethereum Foundation’s dAI team.
The standard was proposed on February 25, 2026, by a team that includes Davide Crapis and Bryan Lim. It defines a protocol for onchain agent commerce, essentially giving autonomous AI agents a structured way to do business with each other on Ethereum.
What ERC-8183 actually does
ERC-8183 introduces what it calls a “Job” primitive. Each Job comes with an escrowed budget, meaning the funds are locked up front so neither party can rug the other.
The standard defines a four-state lifecycle for every Job: Open, Funded, Submitted, and Terminal. The framework also includes evaluator attestation, which means a neutral third party (or mechanism) verifies that the work meets the requirements before payment is released.
ERC-8183 is designed to integrate with ERC-8004, a separate proposal focused on agent reputation. Together, they form a two-layer system: one for doing the work, one for tracking who does good work.
Why Virtuals Protocol is involved
Virtuals Protocol co-developed the ERC-8183 standard alongside the Ethereum Foundation. The project has previously managed over $3 million in agent transactions without escrow, operating in the exact pain-point space that ERC-8183 is designed to solve.
The platform currently hosts a community of over 17,000 agents and has reported revenue of $39.5 million.
What this means for investors
The VIRTUAL token is trading at approximately $0.70 as of late March 2026, with a market capitalization of roughly $457 million.
If ERC-8183 gains adoption as the default standard for AI agent commerce on Ethereum, Virtuals Protocol, as a co-developer with an existing agent ecosystem, would be among the first to benefit. Projects like Autonolas and Fetch.ai are also building their own frameworks for agent interaction, making developer adoption the key variable to watch.
