Sentinel

Inspiration

We understand that navigating finances can be difficult, so we made a way to stream line the process of managing finances for individuals and businesses.

As AI becomes increasingly popular, we imagine a future where people rely on AI agents to manage and spend money for them. That led us to create Sentinel, a way to financially navigate that future.

What it does

Sentinel sits between AI agents and a user's money.

Instead of giving an agent unrestricted access to an account, users create spending policies that guide what agents are allowed to do.

Users can:

Set a protected balance that agents can never spend below.
Connect multiple AI agents to the same financial system.
Give each agent its own spending limit.
Review proposed and completed transactions.
See why transactions were approved or rejected.

For example, a user could have a Necessities Agent with the highest priority, a Travel Agent, a Shopping Agent, and an Entertainment Agent.

When an agent requests a transaction, Sentinel evaluates the request against the user's policies before allowing it to continue.

The goal is simple: let AI act autonomously without giving it unlimited financial authority.

How we built it

We designed Sentinel around three main pieces: the user dashboard, the agent layer, and the policy engine.

The dashboard gives users a central place to see their available balance, protected funds, connected agents, priorities, spending limits, and transaction history.

Each agent is treated as an independent actor with its own permissions and priority. When an agent wants to make a purchase, it sends a spending request to Sentinel rather than directly accessing the user's funds.

The policy engine then evaluates that request. It checks things such as the requested amount, the agent's spending limit, its priority, the remaining balance, and the user's protected balance.

Only after passing those checks is the request approved.

We also focused heavily on the interface. We built the UI around clear balances, simple controls, agent cards, and visual explanations of each decision.

Challenges we ran into

There were many features that we tried to implement such as integration with other APIs. However, at some point, we realized that our scope was too big and that we would not be able to implement all of our ideas into our project. We then cut down the size of the project to best flesh out our most important features within the time constraints.

Accomplishments that we're proud of

We are most proud of how we were able to create an entire full-stack application. Being the first hackathon for everyone on the team, we came into this not knowing fully what to expect, but once our app started being built, we were very happy seeing what we were able to create.

What we learned

We learned how to effectively develop applications in a time constrained environment. We also learned how to work in a collaborative manner to build the application collectively to maximize our efficiency in building.

What's next for Sentinel

Next, we would like to connect Sentinel to real banking and payment APIs so that policies can control actual transactions.

Eventually, we would also like Sentinel to utilize an API that any AI agent can request permission from before spending money.

As autonomous agents begin navigating more of our digital lives, we believe there needs to be an independent system making sure they are still navigating according to our rules.