Lucra AI
Talk money. Lucra does the rest.
The problem Lucra AI solves
Lucra AI solves a critical gap in the current onchain ecosystem: intuitive, intelligent, and user-friendly crypto finance management.
🧩 The Problem
-
Crypto Payments Are Still Technical & Rigid
- Users need to manually enter wallet addresses, use exact tokens, and go through multiple steps to complete even simple payments.
- Natural-language interactions are nearly nonexistent.
-
No Personalized Finance Tools Onchain
- Web3 lacks tools that help users understand their spending, track transactions, or receive smart suggestions based on usage — unlike Web2 neobanks and apps.
-
Multisend & Split Payments Are Friction-Heavy
- Current tools don’t allow easy multi-party payments from one message.
- Most wallets don't support splitting funds without custom scripts or contracts.
-
Chat + Payments Are Disconnected
- No seamless blend of communication + financial action like “Slack meets your wallet.”
- Group payments, tagging, or chatting around money require external platforms or manual effort.
âś… How Lucra AI Solves This
| Challenge | Lucra AI's Solution |
|---|---|
| Manual, confusing payment UX | Users just talk to Lucra in plain English to send, split, or track funds. AI understands intent and handles logic. |
| No smart spending insights | Lucra acts like a personal CFO — categorizing spending, surfacing reminders, and showing trends from past transactions. |
| Friction in multisend & split | Users can say "Send 10 to @bob and @alice" or "Split 100 three ways" — Lucra auto-detects and executes onchain. |
| Lack of chat-native transactions | Group chats and @tags allow money to flow naturally in conversations — just like chatting, but onchain and instant. |
| Poor accessibility for newcomers | Coinbase smart wallet + gasless UX makes it easy even for crypto novices to onboard, transact, and manage finances. |
đź“Ś In short:
Lucra AI replaces forms, wallet toggles, and technical steps with one powerful interface: conversation.
It bridges the usability of ChatGPT with the power of Base's onchain rails — and delivers the first real “Web3-native finance assistant” experience.
mainnet transaction - 0xe83adfea27f20d9b3419033f23bd29c48c99dcd1d2143cbf193a229679f1244e
Challenges we ran into
đź› Challenges We Ran Into
🔄 Challenge 1: Interpreting Natural Language into Onchain Actions
Problem:
Parsing flexible, ambiguous chat commands like “split 10 bucks between Bob and Alice” or “send 5 to Jack if he’s online” into structured transaction logic (token, recipient, amount) was non-trivial.
Solution:
We trained and fine-tuned a GPT-4-turbo-based command parser with a JSON schema constraint and fallback heuristics. We also implemented a keyword-suggestion chip bar (like ChatGPT's follow-up suggestions) to guide ambiguous inputs toward deterministic actions.
⚖️ Challenge 2: Multi-Recipient, Gasless, Cross-Call Execution
Problem:
Executing multiple token transfers in a single sponsored transaction while resolving
@handles
to wallets, calculating equal splits, and dynamically generating calldata posed contract and bundler-level challenges.Solution:
We deployed a multicall-capable smart contract that takes an array of
recipient-amount
pairs. Basename handles are resolved off-chain with caching via AgentKit, and Coinbase Paymaster is used for gasless execution via Smart Wallet.🧾 Challenge 3: Real-Time Transaction History & Summarization
Problem:
Retrieving historical onchain transactions and categorizing them by purpose (e.g., donation, refund, payroll) in real time without flooding the frontend was tricky.
Solution:
We indexed each user’s wallet using an internal Graph node and enriched each transaction with GPT-powered summaries, emoji-tagged categories, and recipient basenames. This data was cached and surfaced via a performant REST API.
🧠Challenge 4: Intelligent Prompt Suggestions & UI Sync
Problem:
Maintaining context-aware suggestions (like “Want to split the bill with @group?”) across multiple chat contexts caused stale UI state and memory bloat.
Solution:
We integrated Zustand for global state management and coupled our suggestion engine with the GPT parser so that every intent path included next-step suggestions as part of the output schema.
đź”’ Challenge 5: Security & Transaction Previewing
Problem:
Users were concerned about blindly signing smart wallet transactions parsed from chat.
Solution:
We built a preview engine that converts parsed intent into a human-readable preview (“You’re sending 5 USDC to @alice.base, gas is 0.00 ETH”), which must be confirmed before signing.
Lucra AI tackles both the technical complexity and the human friction of making onchain interactions feel intelligent, natural, and powerful.
Let me know if you’d like a downloadable one-pager or Notion-ready project doc for this!
Tracks Applied (2)
AI
Consumer
Cheer Project
Cheering for a project means supporting a project you like with as little as 0.0025 ETH. Right now, you can Cheer using ETH on Arbitrum, Optimism and Base.