Agentica

Problem Statement

Agentica is an advanced AI-powered treasury management system designed to automate crypto trading. The primary goal of Agentica is to provide autonomous trading that allows everyone, regardless of their experience or understanding of crypto markets, to benefit from highly informed trading decisions made by AI agents.Agentica leverages Eliza OS to autonomously monitor markets, track price movements, and execute trades for users, all without requiring manual intervention.Users select their strategy (Bitcoin DCA, Base Momentum, or Vitalik Signal), deposit tokens from any chain, and their dedicated Eliza AI agent handles everything.The AI's analysis is driven by several crucial data sources like user input custom strategy parameters and risk preferences.On-Chain Data using The Graph AMP, Agentica monitors decentralized exchange activity, such as Uniswap V3 trades, whale movements, and token liquidity. This on-chain data provides real-time insights into market activity, which the AI uses to make data-driven decisions. Started in AMP playground, explored Uniswap V3 Base schema, and tested SQL queries interactively. It gives the recent swap events and extract unique senders.The Zapper API provides the AI with real-time portfolio tracking, allowing it to understand the user's holdings and monitor the user's current positions in DeFi protocols. This allows the AI to make intelligent decisions regarding asset allocations, ensuring that users' portfolios are optimized for growth.The AI agent also pulls in world news from major crypto sources. By analyzing these signals, Agentica can predict market trends and adjust trades accordingly.Once the AI has analyzed the necessary data, it executes trades autonomously using gas-optimized smart accounts on Base.The AI makes trading decisions in real-time, executing swaps when conditions are met, explaining every action through conversational chat. Deposit once from any chain, your agent trades 24/7 on Base with gas-optimized smart accounts.

Solution

User Onboarding and Wallet Management: Coinbase Embedded WalletsThe user journey begins with Coinbase Embedded Wallets, allowing even non-native crypto users to easily sign in and get started without needing to manage seed phrases. Coinbase’s OAuth service handles the authentication and provides gas-optimized smart wallets for users once they deposit USDC. This wallet infrastructure is crucial because it allows users to store and manage their funds securely on-chain, while the system automates trading and portfolio management.Upon sign-in, users provide preferences for how they want their trading strategy to work. These preferences dictate the behavior of their AI trading agent such as which strategy to follow (e.g., Bitcoin DCA, Momentum-based, or following Vitalik’s trades) and their risk tolerance.AI Framework: Eliza OSEliza OS is an autonomous decision-making framework that powers the trading agents. Each user is paired with a dedicated AI agent that follows the user’s chosen strategy and trading parameters. Eliza OS is connected to the platform via Socket.IO, which provides a real-time conversational interface so users can see what their AI agent is thinking and the reasoning behind each trade. This transparent, chat-like interface ensures users stay informed about the decisions made by the AI agent, fostering trust in the system.The AI agents use custom prompts to determine the strategy they follow, whether it’s buying on dips with Bitcoin DCA or tracking social signals from crypto influencers like Vitalik Buterin. These strategies are then executed autonomously based on real-time data, with the AI adjusting its approach as market conditions evolve.Data Sources and Integration: Graph AMPThe Graph AMP plays a critical role in providing on-chain data, which Agentica uses to identify significant events such as whale movements, volume spikes, and trending tokens on Uniswap V3 on the base mainnet. Using the We also are using the This data is queried using SQL-like queries from AMP subgraphs, allowing Agentica to track trades and liquidity movements across decentralized exchanges in real time.Fetch 200 most recent swaps from the Uniswap V3 using queries like :SELECT * FROM event__swap ORDER BY _block_num DESC.Used binary address filtering likeWHERE event.sender = x'...'(PostgreSQL hex literals for indexed lookups) and data arrives in batches via async iteration. Used this data to feed to the AI models about each wallets transaction details.Coinbase Embedded Wallets are pulled from key sources like Twitter and Farcaster, ensuring the AI is aware of sentiment shifts in the market. This information is used by the AI to make timely adjustments to the user’s portfolio based on market-moving events, whether it’s a government regulation or major market adoption.The Zapper API is used to track users' portfolios, providing real-time data on their token ****holdings, DeFi positions, and NFTs. This data is essential for the AI agent to understand a user's current portfolio and make intelligent trading decisions. It also allows Agentica to ensure that the user's portfolio is appropriately allocated according to the strategy and risk preferences they have set.Transaction Execution and Gas Optimization:Trades are executed using Base L2, a Layer 2 network designed for low-cost execution and high-frequency trading. This ensures that Agentica can execute trades efficiently and with minimal transaction fees. Using the Trade API, the AI swaps between the assets and tokens on the Base and Ethereum chains.The platform uses Coinbase CDP React to deploy non-custodial, gas-optimized smart wallets for each user. These wallets handle all transactions and ensure that the user retains control over their assets while the AI executes trades in the background. Additionally, Daimo Pay SDK facilitates cross-chain USDC deposits, enabling users to deposit assets from chains like Ethereum, Polygon, or Arbitrum, which are converted to USDC on Base.Frontend and UI:The frontend is built with Next.js 15 and TypeScript, offering a server-side rendered architecture for fast performance. The UI uses Tailwind CSS for responsive design and Recharts for visualizing portfolio data. Real-time updates are managed via Socket.IO, which ensures that the user's agent's actions, market data, and portfolio status are reflected in the UI instantly.On the UI side, the user is presented with a clean and intuitive dashboard that showcases their portfolio, the current state of the market, and the actions their AI agent is taking. The platform provides a conversational chat interface that explains trade decisions, making it clear and transparent for users.The particularly hacky thing in our project was figuring out how to make AI agent rooms with multiple agents having multiple strategies.