Building the Onchain Agent Infrastructure for Autonomous Web3

Building the Onchain Agent Infrastructure for Autonomous Web3

As the integration of artificial intelligence (AI) and blockchain technology continues to evolve, the potential for creating autonomous web3 environments is becoming increasingly tangible. According to a blog post by The Graph, autonomous agents need a structured approach consisting of identity management, environmental awareness, and economic agency to function effectively in this new landscape.

Identity: The Key to Trustworthy Interactions

Before an AI-driven agent can perform transactions, it must have a reliable identity recognizable in the blockchain ecosystem. Traditional methods, such as granting an agent access to a user’s private key, raise security concerns and complicate the distinction between human and AI actions. To mitigate these issues, **ERC-8004**, termed the Trustless Agents standard by The Graph, introduces three onchain registries that encompass identity, reputation, and validation.

This infrastructure allows agents to establish verified identities and interact without needing prior trust among parties. Additionally, it supports account abstraction, enabling the development of clearly defined permissions for actions like trading limited amounts daily, ensuring comprehensive security control over user funds.

Data: Enabling AI Understanding of Blockchain

Beyond identity, autonomous agents require accurate data to navigate the blockchain effectively. While the blockchain offers vast amounts of transparent information, it is often dispersed across various transactions and smart contracts, making it challenging for an AI to access the relevant data efficiently.

The Graph’s Subgraphs address this issue by organizing blockchain information into structured datasets, which can be accessed through the Model Context Protocol (MCP). This setup bridges the gap between complex blockchain data and the AI’s needs, allowing agents to retrieve and process information without sifting through extensive raw data.

Economic Agency: Facilitating Transactions with Autonomy

The final layer in the onchain agent infrastructure is equipping agents with the means to engage in economic transactions autonomously. Current payment systems are designed for human users, necessitating manual input for approvals and transactions. To facilitate automated payments, the emerging **x402** standard recasts HTTP’s “402 Payment Required” code, allowing services to request payment as part of data queries directly.

The Graph has integrated x402 into its Subgraph Gateways, enabling agents to pay for queries in USDC without traditional account setups. This innovation streamlines transactions, enabling machine-to-machine commerce with automated micropayments that cater to high volumes of small interactions.

Integrating the Layers: The Agent Loop

When combined, these fundamental layers present a clear framework for the onchain agent infrastructure. An agent can establish its identity, interpret structured data to understand the environment, and autonomously execute transactions without human intervention, effectively participating in what’s termed the “agent loop.”
By leveraging ERC-8004 for identity and accountability, Subgraphs and MCP for contextual understanding, and x402 for seamless payments, these technologies represent a shift toward a more decentralized web where autonomous software not only exists but thrives.

The Graph currently provides this blockchain data infrastructure across 60 networks and has demonstrated its capability by processing over 1.27 trillion queries for more than 75,000 projects. The emergence of autonomous agents marks a promising evolution in how blockchain technology could serve not just humans but also the software itself, potentially laying the groundwork for a machine-to-machine economy at internet scale.