Near Protocol
Useful Agents Hackathon 2025 San Francisco
past
14 Mar, 25 - 15 Mar, 25
The Useful Agents Hackathon was a 24-hour sprint held in San Francisco, United States, co-organized by NEAR AI, SF Compute, and Electric Capital on March 14 - 15, 2025. Designed for builders to develop practical AI agents that reliably execute tasks and meet concrete performance metrics, the event notably did not require projects to interact with crypto or a blockchain. The competition offered a total prize pool of $20,000, alongside dedicated go-to-market support and free inference provided to participating developers. Over 100 developers participated, with top honors awarded for innovations in social media management, due diligence, and anti-bot event discovery.
Near Protocol - Useful Agents Hackathon 2025 San Francisco Overview
How to join Near Protocol - Useful Agents Hackathon 2025 San Francisco?
The event framework required teams of 1 to 4 members to build and host agents exclusively on the Agent Hub. Participants utilized open-weight models running confidentially via Intel TDX and NVIDIA GPU TEEs. A key requirement was submitting a public GitHub repository complete with quantitative benchmarks, a demo video, and reproducible tests. The hackathon kicked off with a panel featuring notable industry figures, including Avichal Garg from Electric Capital and Illia Polosukin from NEAR.
The total prize pool amounted to $20,000, distributed across general awards and sponsor bounties. General prizes included $5,000 for first place, $3,000 for second, and $2,000 for third place. Additional specialized tracks were sponsored by ecosystem partners like Coinbase CDP, Phala Network, and Nevermined, offering specialized bounties ranging from $500 to $5,000 in credits. Winners were officially announced on March 15, 2025, with top projects prioritizing quantifiable AI utility over blockchain integration.
Step-by-step Guide
Register for the event on the Luma portal to secure participation.
Form a team of up to four members and conceptualize a practical AI agent.
Integrate at least one AI component, such as LLMs or ML models, into the project architecture.
Deploy and host the finished agent on the Agent Hub platform.
Publish a comprehensive GitHub repository featuring quantitative benchmarks, reproducible tests, and a demo video.
Present the final project during the in-person judging session in San Francisco.