Building RecallPay: A Money Planner That Remembers What Matters
Money is rarely meant for just one thing. It might need to cover monthly bills, a birthday gift, a date, and a savings goal—all while leaving room for unexpected expenses.

I built RecallPay for the Tatum x Walrus Total Recall hackathon to explore a simple question:

What if a money planner could remember what matters to you and use that context when helping you adjust your plans?

Making room for real life
RecallPay brings monthly expenses, future plans, savings, and reminders into one interface.

A user can enter a planning balance, assign amounts to expenses and goals, and log an unexpected purchase. The app then shows how much is left to plan with.

The prototype uses fictional starting data. Its allocations are virtual: RecallPay does not hold or move funds, pay bills, or cancel subscriptions. Reminders currently appear inside the app.

Why memory matters
Imagine a savings assistant suggests cutting spending by 30%. That might sound reasonable until the user explains that they tried a similar cut before and found it too restrictive.

Without memory, the assistant might repeat the same suggestion later. With memory, it can take that feedback into account.

For our demo, RecallPay uses Walrus Memory on Mainnet to store and retrieve fictional financial preferences. We wrote ten synthetic memories, verified the write jobs, and retrieved memories with their blob IDs.

The live app displays a comparison between generic advice and advice informed by recalled memory.

Closing the feedback loop
We added a protected demo flow where the operator can save one of two fictional responses: “Too strict” or “Helpful.”

When the app recalls “Too strict” in a later session, its recommendation changes to suggest a smaller weekly contribution. If the feedback was “Helpful,” the demo can retain the gradual approach.

The app shows the recalled memory and blob ID so the change is visible rather than hidden behind a claim that the product “learns.”

The current implementation uses explicit rules, not a model that retrains itself. It also does not yet learn from arbitrary conversations. The shared demo is not intended for real personal financial information.

Connecting Sui through Tatum
RecallPay also includes a read-only Sui wallet balance lookup powered by Tatum’s Sui Mainnet gRPC endpoint.

We first tested the connection in a GitHub Actions workflow, then connected it to the live app and verified a lookup using a public test address.

The lookup does not sign a user in, transfer SUI, or automatically add the wallet balance to the naira planning balance.

What we learned
Building RecallPay showed me that persistent memory is most useful when it can change a decision and explain why.

It also made the responsibilities around financial data clear. A convincing demo needs more than a successful API call: it needs visible evidence, careful handling of credentials, protection against unrestricted writes, and honest explanations of what the product does and does not do.

What comes next
The next version would focus on proper user authentication, separate memory spaces for each user, stronger access controls and rate limits, and privacy safeguards.

I would also explore opt-in reminders outside the app, while keeping users in control of whether bills are paid.

Try RecallPay
Live demo: https://recallpay.onrender.com/
GitHub repository: https://github.com/Babytech44/recallpay
Walrus memory test: https://github.com/Babytech44/recallpay/actions/workflows/walrus-memory-test.yml
Tatum Sui gRPC test: https://github.com/Babytech44/recallpay/actions/workflows/tatum-grpc-test.yml
Built for the Tatum x Walrus Total Recall hackathon.