PACT
The_whisperer's Team
Problem Statement
Today's AI agents can chat but can't transact safely. When humans want to use AI to procure services — hiring a contractor, purchasing data, commissioning research — there's no infrastructure for agents to move money, verify deliverables, and enforce agreements onchain with proper guardrails. The current options are dangerous: either give an AI agent unrestricted access to your wallet (reckless), or manually execute every step yourself (defeats the purpose of automation). There's no middle ground where agents can spend within enforced limits, counterparties can be discovered and verified trustlessly, agreements are immutable, deliverables are verified before payment, and sensitive artifacts stay encrypted. PACT solves this by introducing mandate contracts (smart-contract-enforced spend caps per deal), Lit Protocol encryption (artifacts only accessible to authorized parties), ERC-8004 identity (verifiable onchain reputation), Self Protocol ZK proofs (privacy-preserving counterparty vetting), and a full escrow state machine (OPEN → ACCEPTED → DELIVERED → VERIFIED → RELEASED) that ensures funds only move when deliverables are verified.
PACT (Policy-Aware Crypto Transactor) is an autonomous AI agent platform for onchain procurement. It lets AI agents act on behalf of humans — discovering counterparties, negotiating deals, escrowing USDC on Base, encrypting artifacts with Lit Protocol, verifying deliverables with Venice AI, and settling onchain — all governed by human-defined mandate policies with smart-contract-enforced spend caps. Unlike chatbots that can only talk, PACT agents can independently execute the full procurement lifecycle: parse natural-language requests into structured intents, discover counterparties via a multi-source waterfall (seeded directory → ERC-8004 registry → ENS text records), score them with Venice AI, negotiate terms, open USDC escrow deals on Base, encrypt sensitive artifacts with Lit Protocol access conditions, verify deliverables using Venice AI multimodal analysis, and settle funds onchain — all while respecting spending limits, privacy requirements, and approval thresholds defined in a human-controlled mandate. The system uses Vercel Workflow DevKit for durable, crash-resilient deal pipelines; ERC-8004 for verifiable agent identity and onchain reputation feedback; Self Protocol ZK proofs for privacy-preserving identity verification; Storacha/Filecoin for content-addressed encrypted storage; and ENS text records for human-readable agent discovery.
Team
The_whisperer
admin
Increase 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
Exploring
Depends on traction and community interest — open to continuing development if there is demand