Web3 UPI

Making crypto transfers feel as simple as UPI using Gemini-powered transaction guidance.

Comment

Inspiration

Crypto transfers are powerful but intimidating.
For most users, sending money on-chain involves understanding networks, gas fees, address formats, and irreversible mistakes.

In India, UPI made digital payments successful because it removed complexity.
We asked a simple question:

What if sending crypto felt as easy and guided as making a UPI payment?

At the same time, large language models like Gemini are extremely good at explaining complex topics in simple language.
This inspired us to embed Gemini directly into the transaction moment — where mistakes are expensive.

What it does

Web3 UPI is a mobile-first crypto sending interface that simplifies multi-chain transfers.

Users enter a recipient address and the system:

  • Detects the blockchain
  • Adapts visuals and rules
  • Shows amount and fee context
  • Provides AI assistance before the transaction

Instead of acting as a chatbot, Gemini becomes a transaction copilot.
It explains gas fees, flags unusual situations, and refuses unsafe requests like revealing private keys.

How we built it

The application is a client-side progressive web app built using JavaScript, HTML, and CSS.

We created:

  • Address parsing and chain detection logic
  • Modular UI components for send flows
  • Dynamic theming per network
  • Context packaging so Gemini understands what the user is about to do

Gemini receives structured signals (chain, asset, amount, fees) and returns human-readable guidance.

We also implemented layered protections such as:

  • Prompt sanitization
  • Restricted capabilities
  • Response filtering

This ensures the AI can assist but never control funds.

Challenges we ran into

One major challenge was balancing helpfulness vs safety.

If the AI is too open, it risks hallucinating or suggesting unsafe actions.
If it is too strict, it becomes useless.

We had to design prompts, constraints, and guardrails so that Gemini:

  • Educates
  • Warns
  • Clarifies

but never crosses into execution authority.

Another challenge was multi-chain abstraction.
Different networks behave differently, and making them feel uniform required careful UI and logic design.

Accomplishments that we're proud of

We are proud that Gemini is not just added on top, but integrated at the exact moment users need clarity.

The assistant:

  • Understands transaction context
  • Provides real-time explanations
  • Prevents common risky requests
  • Improves user confidence before sending

We also built a polished, responsive experience that feels familiar even to non-crypto users.

What we learned

We learned that AI becomes most powerful when placed inside decision points, not outside them.

Users do not want another chatbot.
They want help exactly when uncertainty happens.

We also learned the importance of strict boundaries in financial environments.
Good AI design is as much about what the system refuses to do as what it can do.

What's next for Web3 UPI

Next steps include:

  • Expanding real execution across more chains
  • Integrating richer risk signals
  • Improving personalization of assistance
  • Exploring deeper agent workflows with strong safety controls

Our long-term vision is to make blockchain transactions understandable and safe for mainstream users.

Built With

Share this project:

Updates