BountySwarm
AI-powered multi-agent system that automates hackathon prize distribution
Five AI agents (Scout, Analyst, Auditor, Compliance, Executor) form a consensus mechanism that evaluates hackathon submissions and releases MNEE stablecoin to winners via Ethereum smart contracts in under 60 seconds.
AI Involvement
BountySwarm is built entirely around a multi-agent AI architecture. Five specialized agents — Scout, Analyst, Auditor, Compliance, and Executor — each evaluate hackathon submissions from a distinct perspective. Scout discovers and categorizes submissions. Analyst evaluates technical depth. Auditor checks for rule violations. Compliance verifies eligibility. Only when all agents reach consensus does Executor trigger the smart contract to release prize funds. This mirrors how human judging panels work and reduces the risk of any single model's blind spots affecting the outcome. No agent can unilaterally pay out or withhold a prize.
Key Technical Decisions
- Multi-agent consensus architecture — no single agent can unilaterally release or block funds
- Specialized agent roles — each agent is prompted for one evaluation lens, reducing hallucination risk
- MNEE stablecoin on Ethereum — programmable, trustless prize distribution via smart contracts
- Smart contract as escrow — funds held until agent consensus is cryptographically confirmed on-chain
- Sub-60 second end-to-end target — performance budget set at hackathon kickoff and met