HooMind

Demo

About

The problem it solves Managing DeFi trades manually is inefficient and fragmented. Users have to switch between multiple platforms like Uniswap for swaps, Compound for staking, various tools for checking prices and manually execute each step. It’s time-consuming, easy to mess up, and not accessible for people who aren't deeply technical. This project automates the entire workflow using an onchain AI agent that users can interact with via XMTP chat, hence extra secure. It acts like a personal DeFi assistant that listens to user commands, monitors the market, and executes trades or staking operations directly onchain and provides concrete analysis dashboard for current investments. What People Can Use It For Traders who want to automate strategies without writing scripts. DeFi users who prefer messaging over dApp interfaces. Anyone who wants to monitor their wallet performance and trade history through a clean portfolio dashboard. Projects or DAOs that want to give their community a tool to automate recurring DeFi actions. How It Makes Things Easier Before this, you'd have to: Go to a price aggregator to check token prices Use a DEX like Uniswap to swap Use a separate platform to stake Manually track performance in spreadsheets or dashboards Now, you can: Just message the agent: "Provide some good investment opportunities/ trending trades across the market" and based on the suggestiong “Swap 100 USDC to ETH if price drops below $3k” The AI agent handles it, confirms via chat, and shows updates on the dashboard The agent: Monitors price feeds in real time Executes transactions from its own wallet Sends updates back to the user through XMTP Displays everything holdings, trade history, staking yield in a portfolio analysis dashboard This removes all friction from onchain trading and portfolio management. Tech Stack Used: XMTP for secure communication Coinbase agentkit for AI bot Uniswap v4 for conducting trades Mongodb for storing and fetching opportunities Compound for staking Langchain for AI integration (Intelligence)

How it's made

Challenges we ran into 1. Agent Wallet Management & Security Giving the agent its own wallet means it needs to hold user funds and execute real transactions. Making sure this wallet is secure—both from external threats and from misbehaving code—was a big challenge. We had to isolate signing logic and make sure the wallet couldn’t accidentally drain itself or be hijacked. 2. Avoiding False Opportunities The AI agent constantly monitors market conditions, but not every price dip or spike is actually a good trade. Filtering out noise from real opportunities was tricky. We had to find a balance between acting fast and not overtrading. This involved tuning how the agent interprets price feeds and reacts to signals.