Compass Investments

– Etherium claims to be the computation layer for autonomous AI systems with the upcoming introduction of ERC-8004, a new standard that gives AI agents portable identifiers and a trusted reputation between different organizations.

This protocol addresses a key trust deficit in machine-to-machine communication by creating blockchain-based discovery and confirmation mechanisms that allow agents to act autonomously without intermediaries.

Davide Krapis, head of AI at the Ethereum Foundation, announced the launch of the protocol in a video message, explaining how the standard enables secure interactions between AIs.

Etherium is uniquely positioned to become the platform that will govern the majority of AI exchanges, Krapis said, adding that between one and two thousand developers have already joined working groups since the specification was published in August 2025.

Etherium is well-positioned to secure and regulate AI interactions.

At the moment, AI agents function as siloed systems, locked within their own ecosystems, unable to communicate across organizational boundaries without predetermined trusted relationships.

OpenAI agents work exclusively within the OpenAI infrastructure, while Google agents only recognize Google’s rules, forming disconnected networks that inhibit their potential for autonomy.

While Google’s A2A protocol and Anthropic’s MCP have established communication standards that allow different AI systems to share data, neither addresses the pressing issue of how agents find trusted partners or authenticate their credentials.

The ERC-8004 fills this niche by providing three basic tools for trust-free agent-to-agent communication.

First, each AI agent is uniquely identified on the blockchain as ERC-721 (FT), which creates verifiable credentials that cannot be tampered with.

Second, the protocol captures reputation scores on the blockchain based on user feedback and successful completion of tasks, similar to Uber driver ratings, but stored permanently on Etherium.

Third, high-risk tasks that require additional verification may use cryptographic proofs, trusted execution environments, or stake-guaranteed verification to ensure accurate agent results.

The standard operates by registering agents with the Identity Registry, which assigns each agent a unique ID and points to a registration file with a list of capabilities, endpoints, and supported protocols, including A2A, MCP, ENS names, and wallet addresses.

The Reputation Registry accumulates feedback from users who have interacted with agents, capturing ratings and optional detailed comments stored in IPFS for permanent access.

4 / Reputation is an onchain asset.

Using scores and custom tags (tag1, tag2), anyone can send, store and summarize reputation signals on the blockchain. Additional advanced event reporting in IPFS allows for deeper off-chain analysis.