AI Agents Begin Executing On-Chain Transactions

Fetch.ai's uAgents can operate across multiple blockchains, including Ethereum, Binance Smart Chain, and any EVM-compatible chain, without being tied to a specific wallet like MetaMask or Keplr. This wallet-agnostic and chain-agnostic approach allows them to execute tasks like trades or governance votes in real-time based on on-chain triggers. These agents are designed to function as autonomous participants within blockchain networks, managing digital assets and facilitating interactions without direct human oversight. The Model Context Protocol (MCP) provides a standardized interface for AI agents to interact with external tools and data, acting as a universal adapter. This allows an agent to decide which tools to use and in what order to complete a task, even enabling human-in-the-loop approvals for certain actions. This protocol is designed to support autonomous AI workflows, moving beyond the reactive nature of previous protocols. Projects like Altered State Machine (ASM) are tokenizing AI agents themselves through NFTs. In the ASM ecosystem, an AI agent consists of a "Brain" NFT, which contains its core learning abilities, and a "Form" NFT that defines its appearance and function. The native ERC-20 token, ASTO, is used for creating and training these agents, participating in play-to-earn games, and governance. On-chain AI agents are not limited to trading; their use cases extend to optimizing DeFi yield farming, detecting fraud, and automating DAO governance. For instance, an agent can monitor liquidity pools across multiple chains like Ethereum, Solana, and Avalanche simultaneously to execute arbitrage opportunities before they disappear. They can also be programmed to automatically rebalance a portfolio based on predefined volatility thresholds or market momentum. The core of an on-chain agent's architecture consists of a wallet for its on-chain identity and asset management, logic that dictates its decision-making, and an intent or objective it is programmed to achieve. To ensure security and prevent unauthorized actions, these agents can operate through programmable wallets with set guardrails, which allows them to execute transactions without ever having direct access to private keys. This infrastructure is crucial for enabling complex, multi-step automated workflows across different protocols. Looking ahead, the development of "agentic" systems is a key focus, where AI agents can operate with greater autonomy. Frameworks like Virtuals Protocol are exploring models where AI agents can own assets, participate in governance, and generate revenue, with the agent's performance directly influencing its tokenized value. This creates economic incentives for developers to optimize agent performance.