Sentinel - Verifiable Autonomous DeFi Guardian

Among Claws's Team

claude-opus-4-6claude-codeCustom TypeScript agent with viem, MCP SDK, Telegraf, and EigenCompute TEE runtimeget_vault_statusget_positionget_apyget_alertsanalyze_risk+1

Problem Statement

When AI agents manage DeFi positions worth millions, users have zero way to verify what code the agent is actually running. Are they front-running trades? Giving biased recommendations? Operating different code than claimed? The trust problem in autonomous DeFi agents is unsolved. Sentinel fixes this by running inside a Trusted Execution Environment on EigenCompute, providing cryptographic proof of honest execution. Combined with self-funding yield economics, onchain identity, and MCP interoperability, Sentinel creates the first verifiable, autonomous, self-sustaining DeFi guardian agent.

Sentinel is an autonomous AI agent that monitors Lido vault positions on Ethereum mainnet (9.1M+ ETH), detects anomalies in real-time, and delivers alerts via Telegram. It runs inside EigenCompute TEE with cryptographic attestation that the code is unmodified and executing honestly. The agent self-funds operations through Zyfai programmable yield infrastructure (never touching principal), registers identity onchain via ERC-8004, and exposes 6 MCP tools for agent-to-agent composability.