Inspir## Inspiration

AI agents are becoming capable of managing money, calling contracts, and making autonomous decisions. However, users still lack a simple way to define spending boundaries and recover when an agent behaves unexpectedly.

We built SentinelVault to make agent autonomy safer and more understandable.

What it does

SentinelVault is a programmable financial safety layer for AI agents. Users can define daily budgets, per-action limits, approved chains, and allowed contracts.

Before an action is executed, SentinelVault simulates the request, calculates a risk score, checks the active policies, and decides whether the request should be allowed, reviewed, or blocked.

Users can also freeze the vault, update policies, approve pending requests, and review a complete audit history.

How we built it

We built a responsive web application that simulates an AI agent treasury workflow without requiring real funds or a wallet connection.

The interface includes a vault control room, policy builder, transaction simulator, risk explanation, approval flow, freeze and recovery controls, and audit history.

The demo focuses on the user experience and safety logic behind programmable agent wallets.

Challenges we faced

The main challenge was making a complex security workflow understandable at a glance. We designed the product around clear risk scores, visible policy decisions, and reversible user controls.

Accomplishments

We created a working end-to-end demo where users can configure policies, simulate agent transactions, approve or block actions, freeze the vault, and inspect the resulting audit trail.

What's next

The next step would be connecting SentinelVault to real smart accounts, on-chain transaction simulation, wallet signatures, and verifiable policy receipts.ation

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for SentinelVault

Built With

Share this project:

Updates

Submission history