

Harbor x Novita Agent Sandbox Hackathon
Build and benchmark AI agents on public Terminal-Bench 2.1 tracks using Harbor and Novita Agent Sandbox.
This is a fully remote hackathon for AI agent builders. Run your agent on public Terminal-Bench 2.1 tracks using Harbor and Novita Agent Sandbox, publish your Harbor Hub result, and compete on the live leaderboard for Novita credits.
No team is required. Bring your own agent, choose a track, run it in Novita Sandbox, and submit your public Harbor Hub job link in Discord.
🏆 $2,800 in total hackathon credits
🌎 Fully remote
⏰ Build & submit by July 3, 11:59 PM PT
🧪 4 public TB2.1 tracks
📊 Live leaderboard
How to Participate
Pick one of the Novita TB2.1 track datasets
Run your AI agent in Novita Agent Sandbox with Harbor
Upload your Harbor job publicly
Submit the public Harbor Hub job link in Discord:
Join Discord Server
Optional: Post your run on X with @novita_labs, #NovitaSandbox, and #Harbor for a chance to be reshared.
Timeline
Build & Submit: Ends July 3, 11:59 PM PT
Review Period: July 6 - July 8
Winners Announced: July 9 - July 10
Prize Pool
Each track winner: $500 Sandbox credits + $200 LLM credits
Total hackathon credits: $2,800
For AI-native startups: If you’re building AI agents, dev tools, evals, or sandbox-heavy products beyond the hackathon, the Novita Startup Program offers up to $10,000 in cloud credits.
Join the hackathon and start building: https://events.novita.ai/