OuiBank - Financial Infrastructure for Autonomous Agents

Monitoring Overview
Monitoring with Agent Details
Reporting Dashboard with Transaction Details
Agent Creation

Inspiration

We think the next big shift is not just AI agents that can think, but AI agents that can transact.

A lot of people are building agent workflows right now, but most agents still hit a wall the moment money is involved. They can suggest, plan, and search, but they cannot safely spend on behalf of a user in a scalable way.

The core question that inspired us was:

What does banking infrastructure look like when the customer is not just a human, but thousands of autonomous agents?

We wanted to build the financial control layer for that future:

wallets for agents
spending permissions and policies
transaction visibility and auditability
efficient funding architecture for many sub-agents

That became OuiBank.

What it does

OuiBank is a prototype for financial infrastructure for autonomous agents.

It lets a user:

Configure an agent wallet
Set a budget, per-transaction limit, daily spend limit, allowed categories, approval thresholds, and autonomy level.
Onboard by voice or manual controls
The wallet setup can be done through a voice onboarding flow (ElevenLabs-style interaction) or through a clean control panel.
Watch agent execution visually
Our main interface visualizes how the main agent coordinates with sub-agents and services, checks policy, chooses payment rails, and executes transactions.
Audit all transactions
We added a reporting dashboard with a transaction database view (mock data for now), so users can review and filter all agent activity.

Key architecture idea

Instead of funding every single agent wallet individually from fiat, we use a distributor wallet + sub-wallet model.
This allows cheaper internal transfers (especially on Solana) and makes the system more scalable when many agents are active.

In short:

Fund once -> distribute cheaply -> control globally -> audit everything

How we built it

We built OuiBank as a frontend-heavy hackathon prototype focused on the user experience and system logic.

Main components

Agent orchestration visualization layer
A live interface showing agent communication, policy checks, and payment flow
Agent wallet configuration modal
A popup onboarding flow for setting budgets, rules, and permissions
Voice onboarding concept
Designed for integration with an ElevenLabs voice agent that can auto-fill the configuration form
Reporting dashboard
A separate dashboard page with a transaction database and filters (currently mock data)

Product design decisions

We intentionally focused on:

making autonomy feel safe (policy-based controls)
making transactions feel observable (message feed + transaction state)
making the system feel judge-friendly (clear UI, concrete scenarios, strong visual metaphor)

Tech approach

We used a modern web stack to move fast:

Next.js for the app framework
React for UI architecture
Tailwind CSS for styling
Framer Motion for UI animations and microinteractions

We also designed the state structure so mock transaction data can later be replaced with real wallet / payment data.

Challenges we ran into

This project is ambitious for a hackathon because it combines:

agent UX
fintech UX
policy logic
wallet architecture
visualization
voice onboarding

Our biggest challenges were:

1. Making an abstract idea understandable fast

"AI bank for agents" is exciting, but also abstract.
We had to design the UI so a judge can understand the story in seconds:

configure the wallet
give the agent limits
watch the agent execute
inspect the audit trail

2. Balancing "wow" with trust

We wanted the product to feel futuristic, but not like a toy.
That meant building a polished visual experience while still emphasizing:

limits
approvals
rules
reporting

3. Time constraints

We had to prioritize what would be most impactful in the hackathon window:

a strong end-to-end flow
mock but realistic transaction data
clean interfaces that communicate the architecture

Not every part is fully production-connected yet, but the core product logic and interaction model are there.

Accomplishments that we're proud of

We are especially proud of:

A clear product narrative for a new category
OuiBank makes "agent-native banking" tangible and easy to grasp.
Wallet configuration before autonomy
The setup flow makes trust and user control the starting point, not an afterthought.
Voice onboarding concept
The idea that a user can simply speak their policy and budget to configure an agent wallet feels very natural and powerful.
Distributor wallet + sub-wallet architecture
This is a meaningful systems idea, not just UI. It gives the project a strong scalability story.
Reporting / audit dashboard
We added a dedicated reporting view to reinforce transparency and governance.
A polished hackathon-ready UX
Even as a prototype, the project feels like a real product direction.

What we learned

We learned that the hardest part of agent products is not just intelligence, it is trust design.

In practice, that means:

users need controls before execution
autonomy needs policy boundaries
payment flows must be visible
audit trails are essential

We also learned that a strong visual system helps people understand complex infrastructure much faster.
By turning policy, rails, and transactions into a visible experience, the product becomes much easier to demo and discuss.

Finally, we learned that voice is a great fit for onboarding agent permissions. Saying: "Give this agent 500 euros, max 80 per transaction, travel and groceries only" is much more natural than clicking through 10 settings.