Lookout: Agent Trust Protocol
Lookout's Team
Problem Statement
AI agents are increasingly transacting autonomously - paying for services, cooperating with other agents, managing treasuries. But there is no trust layer: an agent has no way to verify whether the counterparty agent it is about to send funds to has a history of failed transactions, suspicious behavior, or is brand new with zero history. Lookout solves this with onchain behavioral scoring: before your agent pays or cooperates, it checks TrustScore.
The credit score for AI agents. Lookout audits onchain behavior from public transactions, integrates ZK identity via Self Protocol, and writes composable TrustScores (0-100) that any agent, dApp, or human can query before transacting. No target cooperation required - credit bureau model. Deployed on Celo and Base mainnets.
Team
Lookout
admin
MandateIncrease your chances to win
- ›Most agents in the hackathon are exposed to prompt injection
- ›This might cause overspending and loss of funds
- ›Security is a crucial part of the hackathon
Share on X
Tell the world about this project
Tracks
Intention
Plans to continue
Lookout is designed as long-term infrastructure. The trust problem for AI agents only gets harder as agent-to-agent transactions scale. Plan: expand to more chains, add reputation portability via ERC-8004, build SDK so any agent can integrate TrustScore checks natively.