Investre
The Farcaster-native AI trading bot.
About
The problem it solves Crypto trading is fragmented and full of friction. Users discover "alpha" and trading ideas on social platforms like Farcaster but must switch to complex DEX interfaces or other apps to act on them. This delay between discovery and execution can mean missed opportunities. Furthermore, community engagement tools like token giveaways are often manual, inefficient, and lack on-chain transparency. Investre solves this by embedding a powerful, MEV-protected trading bot directly into the Farcaster social layer. What people can use it for: - Seamless Social Trading: Users can execute trades using simple, natural language commands in a Farcaster cast (e.g., "@investre buy $100 of DEGEN"). There's no need to leave the app. - - Automated Social-Triggered Investing (Auto-Buy): This is our killer feature. Users can configure the bot to automatically purchase tokens mentioned in casts that they like or recast. This turns passive social engagement into an active, automated investment strategy. - - Gas-Efficient Community Rewards: Through on-chain Raindrops and Giveaways, creators and communities can distribute tokens to their audience transparently and efficiently. Our RaindropEscrow smart contract uses batch transfers to save 60-80% on gas fees compared to individual transfers. - - Simplified Fiat Off-Ramp: Users can convert their USDC/USDT holdings directly to six different fiat currencies (NGN, KES, USD, GBP, EUR, CAD) via our integration with Paycrest and Coinbase. - AI-Powered Portfolio Insights: Beyond trading, users can request an analysis of their Farcaster history to get fun, speculative insights into their personality, or get a serious analysis of their crypto portfolio's performance. Manage Advanced Orders Simply: Set up sophisticated orders like Take Profit/Stop Loss and gasless Limit Orders using plain English, without ever leaving the Farcaster app. Engage the Community with Crypto: Create token "Raindrops" (batch distributions) or Giveaways to reward followers, powered by gas-efficient smart contracts. Smart Contract Address(es) Our project is powered by several onchain contracts on the Base network: RaindropEscrow Contract: 0xdaA5Ce7E300d628Be05969b2D701d995e36861Bd This contract locks funds for scheduled "Raindrop" events and performs highly efficient, gas-saving batch transfers to all participants. Auto-Buy Contract: 0xAaB671d2562c57156b7C5c77D2a40cB1B9C00186 The optimized smart contract that enables the gasless "auto-buy on like/recast" feature for users with external wallets. Paycrest Gateway (Integration): 0xE8bc3B607CfE68F47000E3d200310D49041148Fc We integrate with this third-party contract to power our direct, onchain fiat off-ramp feature. Other Onchain Proof All activity generated by Investre is verifiable onchain, creating a public record of its utility and traction. Bot's Farcaster Profile: farcaster.xyz/investre - Our bot's direct interactions, replies, and cast announcements are public. RaindropEscrow Contract on Basescan: 0xdaA5Ce...61Bd - Every token distribution creates a verifiable onchain transaction history. Auto-Buy Contract on Basescan: 0xAaB67...0186 - All gasless auto-buy trades facilitated by our relayer are public. User Trades: Every trade executed through the bot is a standard DEX swap transaction routed through protocols like OpenOcean or KyberSwap, visible onchain.
How it's made
Challenges we ran into Building a seamless social trading experience presented several significant technical hurdles: Challenge: The Unreliability of Social Token Distribution Problem: Our initial "Raindrop" feature was memory-based and unreliable. If the server restarted, all scheduled distributions were lost. It also only supported a few hardcoded tokens and couldn't handle viral casts without failing. Solution: We completely re-architected the system. We implemented a PostgreSQL database for persistence and deployed the RaindropEscrow.sol smart contract on Base. This contract locks funds at the time of creation and executes a gas-efficient batch transfer in a single, atomic transaction, increasing reliability from ~60% to over 95% and saving up to 80% on gas fees. Challenge: Prohibitive Gas Costs for Micro-transactions Problem: The "Auto-Buy on Like" feature was not economically viable for small trades (e.g., $1) due to high gas fees on Base. Solution: We designed and deployed a hyper-optimized FlashBotAutoBuy_Optimized.sol smart contract. For external wallet users, we integrated an EIP-2612 permit system, allowing them to grant a one-time approval. Our backend relayer then pays the gas for their subsequent auto-buy transactions, creating a truly gasless user experience. Challenge: AI Intent Recognition Failures Problem: Our first AI model was inconsistent and often failed to understand complex user commands, leading to a poor user experience. Solution: We built a resilient, multi-model AI fallback system. Requests are first sent to the most capable model (Claude 3.5 Sonnet), then fall back to GPT-4o, and finally to a fast regex-based pattern matcher for simple commands. This ensures the bot is always responsive and understands user intent with near-100% accuracy. Challenge: Handling Slow, Asynchronous Operations Problem: Features like the AI-powered personality analysis required fetching a user's entire cast history, which could take up to 30 seconds. A user would mention the bot and get no reply for a long time, thinking it was broken. Solution: We implemented an asynchronous job pattern. The bot provides an immediate acknowledgment reply ("🔮 Analyzing your profile...") and then starts a background job. Once the long-running task is complete, the bot posts a second, separate reply with the final result.