Trust scoring for AI agents on EVM chains. You paste an agent's address, we pull its full transaction history from Blockscout, run behavioral analysis across 9 dimensions (activity patterns, fund flows, counterparties, timezone fingerprinting, protocol loyalty), then score it 0 to 100 using an LLM. The score gets published onchain as an attestation via ERC-7506. There's also a Telegram bot and an autonomous scanner that finds new agents from the ERC-8004 and Olas registries every 5 minutes and flags anything suspicious.
Make it possible to verify any onchain AI agent before you trust it with your money. Right now there's no reputation layer for autonomous agents. We want to change that by building the infrastructure that scores, attests, and monitors agent behavior across every major EVM chain. If an agent is moving funds, there should be a public, onchain record of whether it's been trustworthy or not.
AI agents are moving real money onchain right now. Keepers, liquidators, MEV bots, bridge relayers. There are thousands of them across EVM chains and that number is growing fast. But if you interact with one, you have zero way of knowing whether it's legit. No reputation, no track record, no audit trail. You're trusting a wallet address and hoping for the best. One bad agent can drain a pool, front-run your transaction, or rug an entire protocol. The ecosystem has no standard way to evaluate agent trustworthiness before it's too late.
AgentAuditor pulls the full transaction history for any agent address across 6 EVM chains using Blockscout. It runs behavioral analysis across 9 dimensions: activity patterns, fund flows, counterparty relationships, timezone fingerprinting, failed transaction rates, token movements, protocol loyalty, wallet age milestones, and balance trajectory. Then it sends that profile to Venice AI which scores the agent 0 to 100 across four axes (transaction patterns, contract interactions, fund flow, behavioral consistency) and flags anything suspicious. The final score gets written onchain as an attestation via ERC-7506, so anyone can check it programmatically. On top of that there's an autonomous scanner that discovers new agents from ERC-8004 and Olas registries every 5 minutes and alerts via Telegram when something looks off.
We charge per API call for bulk agent audits. Free tier covers individual lookups. Enterprise plans for protocols that want to screen agents before granting permissions
Want to grow through just grant funding
Global
0
