The perfect gift to onboard your loved ones onchain: LabuBank, your friendly crypto companion.
As crypto enthusiasts, we all have wanted to onboard friends and family. However, it seems like the industry's best offerings are hidden behind layers upon layers of technical complexity and UX disaster. All the while, the most hollow and predatory applications are the only things that appeal to everyday users. Despite massive growth in the blockchain industry the question remains: "How are we going to onboard the everyday individual to the things that excite US like stablecoin payments, DeFi yield, and tokenization?"
The answer? Labubus. This collectible toy has sold over $870m in the first half of the year, largely due to speculation on resale value, like many viral crypto schemes. What if your viral speculation toy was also your portal to a new on-chain life? A hardware wallet that enables onboarding, stablecoin payments and natural language portfolio management? The goal of LabuBANK is to be the first step for many in moving on-chain, a silly giftwrapped Trojan horse to an onchain future.
How It's Made:
This project has to juggle a lot for the sake of nice UX. It includes a defi AI agent, pyUSD onboarding, curve swapping, a hardware wallet and more. Here we'll break down each element and how we built it and tied it together.
Frontend App: The frontend app is hosted remotely on a digital ocean VPS and mobile optimized. To connect the webapp with the toy, we use an NFC chip that has the site URL encoded and will bring up the site upon being tapped.
Hardware Wallet: The M5Stack cardputer is an arduino device with wifi and bluetooth connectivity and a full keyboard. It's flexibility allows us to host a simple HTTP server on the cardputer itself. This way we can have an endpoint /send which will sign a json transaction payload with a hardcoded private key and return the raw transaction. It alerts the user with audio upon receiving a signature request, and the user is prompted to press a button to send the signature back.
NFT Contract: Each Labubank is an NFT written in Solidity, written and deployed via Foundry. The individual NFTs are only mintable by the manufacturer of the toys, and users can then set the on-chain name of their NFTs they own to make them their personalized avatars.
pyUSD Onboarding: To onboard pyUSD to users we use the Coinbase CDP but pay out to a wallet generated by the "orchestrator", a backend service that will accept the incoming USDC and use a smart contract to swap for pyUSD on Curve, sending directly back to the user. Instead of the user having to sign multiple transactions, the user simply confirms a single "coinbase transfer", and the backend handles the pyUSD transfer to the users's wallet. There is a trust assumption in having the orchestrator 'temporarily' custody user funds.
A particularly hack friendly tool we used here was "Pocketbase" an all in one backend solution, that hosts an SQLite database and a REST API in order to keep track of fund flows.
LabuBank Conversational Agent: The core agent is built using ASI:One since it has already been pre-trained on cryptocurrency knowledge. This knowledge base can then be extended by our integration with OpenSea's MCP server. This gives us on-chain context for our particular wallet that the user can query and generalized up-to-date information on-chain.
Curve Router deployed at: https://etherscan.io/address/0xc4B88dF36d721bC80F6Ce1e7373E9e0C38598555
LabuBank deployed at: https://etherscan.io/address/0x26E427f68355d97d7FDEb999A07348194D298415
We noticed there's an issue with uploading GitHub repositories. So we will add it here just in case:
Organization: https://github.com/LabuBank https://github.com/LabuBank/LabuBANKCardputerServer- Cardputer Hardware Wallet https://github.com/LabuBank/LabuBANKSmartContracts- NFT Contracts https://github.com/LabuBank/eth-nyc-router Curve Router Contracts https://github.com/LabuBank/eth-nyc-2025 Frontend Code https://github.com/LabuBank/eth-nyc-orchestrator DeFi Backend Code https://github.com/LabuBank/eth-global-2025-agent-mcp AI Backend Code