AgntID Hackathon
Build an AI Agent You'd Actually Trust
Build something useful. Then make it safe enough to trust.
Imagine an AI agent investigating production incidents, accessing sensitive data or performing financial transactions. It has the tools it needs. But what happens when it attempts an action outside its intended task?
Join 25–30 experienced builders for a hands-on hackathon. Build a useful AI agent, then explore how to enforce its access boundaries at runtime using AgntID.
🎯 The Challenge
Build a real-world AI agent, or team of agents, that performs meaningful work using one or more external tools.
Then identify where unrestricted access creates risk. Decide which actions, resources and parameters your agent should be allowed to access, and under what conditions. Use AgntID to enforce those boundaries during execution.
Your final demo should show
What your agent could do before.
What it should actually be allowed to do.
How AgntID enforces that boundary at runtime.
At minimum, demonstrate:
A useful, working AI agent or agentic workflow
One or more external tools
Permitted and restricted actions
AgntID enforcing those boundaries during execution
A clear before-and-after demonstration
💡Need Inspiration?
Build an AI Finance Assistant that retrieves financial data and processes approved transactions, with permissions changing by task.
Build an AI Customer Support Agent that resolves requests while enforcing boundaries around customer records and sensitive actions.
Or bring your own meaningful agentic workflow.
🔍 What We're Looking For
We care about the problem you solve, the agent you build, the access-control challenge you identify and how effectively you use AgntID.
🏆 Prize
A MacBook Neo for the winning team.
🗓️ Schedule
10:00 AM - Check-ins
11:00 AM - Opening
11:15 AM - Technical walkthrough + security session
12:00 PM - Hacking begins
4:00 PM - Hacking ends
4:00 - 4:30 PM - Judging
4:30 - 5:30 PM - Demos
5:30 PM - Closing & Awards
See you at the hackathon!