AI Web3 Agent - Mobile-First On-Chain Assistant

Inspiration

Web3 technology promises to revolutionize finance and digital ownership, but its interfaces remain overwhelmingly complex. New users are confronted with confusing dashboards, technical jargon, gas fees, and long hexadecimal addresses. As someone passionate about making technology accessible, I was inspired to build a bridge between the power of blockchain and the simplicity of natural conversation. I wanted to prove that you don't need to understand smart contracts to interact with them—you should be able to just say what you want to do. The ultimate inspiration came from my mobile-first environment: building this entire project on an Android phone using Termux, to demonstrate that Web3 development can be truly accessible and not require expensive hardware.

What it does

AI Web3 Agent is a conversational AI-powered assistant that translates natural language commands into Web3 actions. Users simply type or speak their intent (e.g., "Swap 10 USDC to ETH" or "Send 5 USDT to 0x123..."), and the agent parses the command to execute the requested on-chain operation. Key capabilities include:

• 💱 Swaps:Convert between tokens like USDC and ETH with estimated rates
• 📤 Transfers:Send tokens to any wallet address with gas fee previews
• 🌉 Bridges: Move assets across chains like Arbitrum and Base
• 💰 Balance Checks: Instantly view wallet holdings across multiple tokens
• 🔗 Wallet Connection:Simulate connecting a Web3 wallet with live status
• 📜 Command History:Track all interactions for easy reference ## How we built it The entire project was built using a unique, resource-constrained mobile-first approach:
• Development Environment:Built entirely on an Android phone using Termux (Linux terminal emulator), proving Web3 development doesn't require a laptop.
• Frontend:Vanilla HTML5, CSS3, and JavaScript for a lightweight, responsive, and zero-dependency experience that works flawlessly on any device.
• AI Agent:A smart command parser using pattern matching and natural language processing to interpret user intent and extract key details (amounts, tokens, addresses, chains).
• Deployment: Hosted as a static site on Vercel, with code versioning on GitHub for a professional workflow.
• Design: Mobile-first UI with dark theme, glassmorphism effects, and gradient accents for a modern, professional look. ## Challenges we ran into
• Mobile Development Constraints:Building on Termux presented challenges with package management, network timeouts, and limited resources. Resolved by using lightweight vanilla JS instead of heavy frameworks.
• Deployment Complexities:Initial attempts with Next.js failed due to Tailwind CSS dependency conflicts on Vercel. Solved by switching to a simple static HTML deployment, proving that simplicity often wins.
• Natural Language Parsing:Creating a robust command parser that handles various phrasings (e.g., "Swap 10 USDC for ETH" vs "Convert 10 USDC to ETH") required careful regex design and extensive testing.
• Network Reliability:Persistent npm timeout issues on Termux required using Personal Access Tokens and alternative registry configurations.
• Time Constraint:Building a complete prototype from scratch on a phone within the hackathon deadline demanded focus on core functionality and MVP approach. ## Accomplishments that we're proud of
• ✅ Working Prototype: A fully functional Web3 AI agent that processes real commands and provides meaningful responses.
• ✅ Mobile-First Development:Proved that a sophisticated Web3 app can be built entirely on an Android phone using Termux, challenging the notion that you need a high-end laptop.
• ✅ Live Deployment:Successfully deployed at https://ai-web3-agent.vercel.app, accessible to anyone instantly.
• ✅ Professional UI:Created a polished, gradient-themed interface that rivals desktop Web3 apps, optimized for mobile viewing.
• ✅ Zero Dependencies:Built with pure HTML/CSS/JS—no frameworks, no bloat, lightning fast.
• ✅ Command History:Implemented persistent history to improve user experience and debugging.
• ✅ Wallet Simulation:Added wallet connect/disconnect with realistic balance displays and network status. ## What we learned
• Simplicity Wins:Complex frameworks aren't always necessary; vanilla JS can build powerful, production-ready apps faster and with fewer bugs.
• Mobile is the Future:The constraints of mobile development (limited resources, smaller screen) force better design decisions and cleaner code.
• Accessibility Matters:Natural language interfaces are the key to mass Web3 adoption—users don't want to learn Solidity, they want to transact.
• Resourcefulness:Building in a constrained environment (Termux) taught creative problem-solving and efficient coding practices.
• Deployment Strategy:Understanding build processes, static vs. dynamic hosting, and when to choose simplicity over complexity is crucial.
• User Experience: Clear, immediate feedback (like gas estimates and transaction previews) builds trust and confidence in Web3 applications. ## What's next for AI Web3 Agent
• Real Blockchain Integration:Connect to actual networks (Base Sepolia, Arbitrum Sepolia) using libraries like ethers.js or viem for live transactions.
• WalletConnect Implementation:Replace simulated wallet connection with real WalletConnect v2 for mobile wallet integration (MetaMask, Trust Wallet, etc.).
• AI Model Integration:Replace pattern matching with actual LLM APIs (Venice AI, OpenAI) for more sophisticated natural language understanding.
• Multi-Chain Support:Expand to support more networks (Ethereum, Polygon, Optimism) with automatic chain switching.
• Voice Commands:Add speech-to-text for hands-free interaction—ideal for mobile users.
• Transaction History:Store and display past on-chain transactions with status tracking.
• DeFi Aggregation:Integrate with DEX aggregators (1inch, Paraswap) for best-price routing.
• PWA Support:Convert to Progressive Web App for native-like mobile experience with offline capabilities.
• User Accounts:Add authentication (email/social) to save preferences and transaction history.
• Education Mode: Include tooltips and explanations to onboard new users to Web3 concepts gradually. Live Demo:https://ai-web3-agent.vercel.app
  GitHub: https://github.com/Sule-Bashir/ai-web3-agent. Built for NTU InnovateX Hackathon 2026 • Track 2: Web3 Applications, AI Agents and Real-World Use Cases • Co-organised by NTU CCTF and SNZ

Built With

• ai-agent
• arbitrum
• base
• blockchain
• css3
• defi
• ethereum
• frontend
• github
• html5
• javascript
• mobile-development
• natural-language-processing
• prototype
• python
• responsive-design
• sepolia
• static-site
• termux
• ui/ux
• vanilla-js
• vercel
• walletconnect
• web3