Turning wallet intent into inspectable actions without taking over the signature.
State
prototype
Context
hackathon / collaboration
Validation
award / live-deployment
Nova's authentic sprint presentation pairs the public landing surface with the agent workspace the team built. It records the completed product idea, not a claim that every historical integration is still operating today.
Outcome
Nova received 1st Notable Mention and 1st Social Media Challenge recognition in the Southeast Asia Lisk Builders Challenge 3.
Exact hackathon recognition for the named team; not a claim of a grant, customer adoption, investment, or production deployment.
Crypto workflows spread one user intention across wallet menus, raw addresses, explorers, market tools, and payment interfaces. Nova explored whether a conversational layer could translate that intention into inspectable reads and prepared actions while leaving final authorization with the user's wallet.
Intended users
EVM wallet users who want natural-language balance, activity, and transaction preparation.
On-chain explorers and traders combining portfolio, counterparty, whale, and trade-cost context.
Freelancers and creators experimenting with fiat-to-crypto payment links.
Contribution
Full-Stack & AI Builder
Nova was a team hackathon build. Wildan owned full-stack and AI-agent work, Galih contributed interface design, and Arga built and hardened multi-chain intelligence. The institutional award roster also names Casta Garneta.
Next.js product and API implementation.
CopilotKit agent behavior and generative UI actions.
Transaction preparation and user-signing boundaries.
Prediction, Paylink, rate-limiting, deployment, and documentation work.
Decisions
Choices that shaped the build.
01
Orchestrate an existing wallet instead of creating custody
Nova could interpret intent and prepare a transaction while the connected EVM wallet remained the only place where the user approved a signature.
The prototype demonstrates agent-assisted execution without claiming that the AI autonomously controls funds.
02
Render purpose-built evidence cards
Balances, counterparties, trade-cost estimates, payments, and transaction proposals need different information hierarchy; a text-only answer would hide the details a user should inspect.
Tool results become reviewable UI, but each card also increases the surface that must keep provenance and loading states honest.
03
Separate the prediction service behind an HTTP boundary
The trade-cost model could be deployed and checked independently from the Next.js application.
The companion service remained live after the sprint, while its estimate must still be labeled as model output rather than a guaranteed execution price.
System flow
From intent to visible evidence.
01
A user opens the chat route and connects an EVM wallet.
02
CopilotKit maps the prompt to a registered balance, analysis, prediction, payment, or transaction action.
03
Read actions call Next.js routes that select chain clients and aggregate explorer or RPC data.
04
Nova renders a dedicated evidence card, form, or transaction review state.
05
A send action validates the proposal and requests a network switch when needed.
06
Only a separate user action opens the connected wallet for final signing.
Evidence
Product and team, in frame.
Nova's team at the Lisk Builders Challenge 3 outcome: 1st Notable Mention and 1st Social Media Challenge. The exact recognition is also documented by Telkom University's institutional report.
At the end of Nova's hackathon sprint, the award boards keep the external result visible while the people around the build remain the center of the frame. Wildan worked as a Full-Stack & AI Builder across the agent, transaction, payment, data, and deployment paths.
Reflection
What remains, and what comes next.
Current boundary
Nova is a hackathon artifact, not a production or custodial wallet.
The public chat runtime was degraded during the 2026-07-11 audit.
Simulated payment completions, fallback results, and synthetic reconciliation records are not valid product proof.
Next improvements
Preserve a canonical, sanitized demo of the completed sprint flow.
Make live, recorded, staging, and simulated states explicit in data and UI.
Restore a reproducible build, lint baseline, automated tests, and CI.
Harden payment validation and webhook authentication before presenting Paylink beyond prototype scope.
