InsightAI

AI Telegram bot turning whale alerts into actionable insights using LI.FI,Uniswap V4,ENS & Chainlink

InsightAI

Created At

HackMoney 2026

Winner of

ENS

ENS - Integrate ENS

Prize Pool

Project Description

Insight AI is an autonomous crypto intelligence agent delivered through Telegram, built for ETHGlobal HackMoney 2026. When a whale moves large amounts of ETH, the bot detects the transaction, resolves the wallet's ENS identity (name, avatar, Twitter, description) via Ethereum mainnet reverse resolution, and queries Chainlink oracle contracts on Sepolia for verified ETH/USD, BTC/USD, and LINK/USD pricing with confidence scores and node counts. The AI agent, powered by Google Gemini, then analyzes Uniswap v4 pool activity to detect whale swap patterns and Hook behavior — identifying LimitOrder, TWAMM, and Dynamic Fee hooks across WETH/USDC, WETH/USDT, and WETH/DAI pools. Simultaneously, it scans cross-chain positions across 6 EVM chains (Ethereum, Arbitrum, Optimism, Polygon, Base, Avalanche) via LI.FI bridge intelligence, tracking bridge preferences across Stargate, Across, Hop, Connext, and Celer to generate bullish/bearish/neutral signals. The agent autonomously evaluates oracle confidence and calculates cost-benefit ratios to decide which premium data APIs to purchase — paying for each one through the x402 micropayment protocol using USDC transfers on Ethereum. An Express.js server exposes four x402-gated endpoints: Etherscan analysis, historical patterns, sentiment analysis, and market impact. Everything — Chainlink oracle data, Uniswap v4 pool health, LI.FI cross-chain movements, ENS whale identity, and purchased API data — is synthesized by Gemini into a single comprehensive analysis report delivered back to the user on Telegram. The project targets three hackathon tracks: Uniswap Foundation's "Agentic Finance" ($5K) for Hook-aware autonomous trading signals, ENS's "Integrate ENS" ($3.5K) for full identity resolution throughout the app, and LI.FI's "Best AI x LI.FI Smart App" ($2K) for cross-chain whale intelligence.

How it's Made

The project is a Turborepo monorepo built with Bun and TypeScript. The Telegram bot runs on Telegraf v4, with Google Gemini (gemini-1.5-flash) powering the autonomous AI agent. On-chain interactions use ethers.js v6 on Ethereum Sepolia — Chainlink price feed contracts are called directly for ETH/USD, BTC/USD, and LINK/USD oracle data. ENS resolution hits Ethereum mainnet via provider.lookupAddress() and resolver.getText() for names, avatars, and social profiles. The Uniswap v4 service monitors WETH pools and detects Hook usage patterns (LimitOrder, TWAMM, Dynamic Fee). LI.FI integration tracks whale positions across 6 EVM chains (Ethereum, Arbitrum, Optimism, Polygon, Base, Avalanche) with bridge intelligence. Premium data is purchased via the x402 micropayment protocol — the agent autonomously sends USDC transfers for each API call. The Express.js x402 server exposes 4 gated endpoints. All data (users, wallets, whale alerts, payments, price snapshots) is stored in PostgreSQL via Prisma ORM hosted on Neon.

background image mobile

Join the mailing list

Get the latest news and updates