#mira $MIRA

AI agents are starting to look less like chatbots and more like economic actors.

They call APIs, move funds, negotiate prices, and increasingly, interact with other agents without a human in the loop. That creates a simple but brutal question: when an agent wants to act on your behalf or take your money, why should you trust it?

ERC‑8004 is Ethereum’s first serious attempt at answering that question. It doesn’t define how agents think or talk; instead, it gives them a shared on‑chain system for identity and reputation. You can think of it as basic public infrastructure for an AI-native marketplace.At the core of ERC‑8004 is an identity registry. Each agent gets an ERC‑721 token that represents its “agent ID,” and that token points to a registration file with metadata: what the agent does, which endpoints it speaks (A2A, MCP, HTTP), which addresses it uses, and which trust features it supports. That turns every agent into a first‑class on‑chain object: discoverable, referenceable, and even transferable like any other NFT.

On top of identity, ERC‑8004 adds a reputation layer. After an interaction, a user or another agent can leave a rating and a link to richer feedback. Those ratings are stored onchain, which means any application can query an agent’s average score, number of reviews, or history of feedback and build its own view of “is this thing worth dealing with?”

ERC‑8004 also defines a validation registry for task‑level checks. An agent can submit a validation request that includes a hash and URI of its output and the address of a validator, which might be another contract, an off‑chain service, or a verification network. The validator does whatever checks it deems appropriate, then posts a response back on‑chain with a score or pass/fail signal and an optional pointer to a proof.