IntelliFi

Project Description This project provides a decentralized, AI-driven portfolio management service composed of two main components: a web-based Agent Marketplace and a proof-of-concept (POC) Agent. The Agent Marketplace allows users to explore various Agents offering different portfolio strategies—ranging from risk profiles to rebalancing frequencies—and subscribe to the one that aligns with their investment goals. Once subscribed, a user can deposit tokens into a portfolio manager smart contract that designates the chosen Agent’s multisig address as the sole authority for managing those funds. This setup ensures that only the authorized Agent can execute transactions, enhancing security and trust.

Through the web interface, users can monitor real-time token balances, view their portfolio’s overall performance, and access transparent activity reports. Each report is published to IPFS, detailing every action the Agent has taken, including which tokens were swapped, the amounts involved, and the rationale behind these decisions. While the Agent has the authority to rebalance or optimize the portfolio, users retain the freedom to withdraw at any time through the website, allowing them to maintain full control over their assets.

Agent (Proof of Concept) The proof-of-concept Agent demonstrates how an AI-driven entity can manage user portfolios by gathering on-chain balances, real-time price data, and historical market information. It compiles all this information into a prompt sent to a Large Language Model (LLM)—via Nillion or OpenAI’s API—that recommends trades, rebalancing actions, or risk assessments. The Agent then generates a detailed report describing its proposed actions and underlying logic, stores this report on IPFS for transparency, and proceeds to execute the corresponding token swaps using Uniswap V3. After completing its cycle of analysis and trades, the Agent pauses until the next scheduled interval or user-initiated check.

Why This Matters By combining AI-driven decision-making with DeFi smart contracts, this project seeks to simplify portfolio management for both novice and experienced investors. Automated optimization reduces the complexity of active DeFi investing, while on-chain and IPFS-based transparency ensures that users can verify every transaction. A multisig setup preserves security by preventing any single entity from misappropriating assets, and the user’s ability to withdraw funds at any time reinforces trust. Ultimately, it is a modular approach that showcases how AI can enhance DeFi portfolio strategies while maintaining user autonomy.

How It’s Made The project leverages multiple technologies and frameworks to create an AI-driven, multi-agent portfolio management solution. Open Autonomy and Open AEA from Autonolas provide the foundational tools for building and coordinating decentralized off-chain agents, while Ethereum-compatible smart contracts secure user funds in a dedicated portfolio manager contract. A Gnosis Safe (or similar multisig) ensures that any funds moved require authorized agent signatures.

When the Agent decides a token swap is necessary, it integrates with Uniswap V3 for trades, drawing price data from sources such as CoinMarketCap and historical performance data from Uniswap or other DEXs. The LLM—accessed via Nillion or OpenAI—analyzes these data points, offering trade suggestions that the Agent then documents in IPFS-hosted reports. The project’s front-end website allows users to deposit tokens, subscribe to preferred Agents, and track their holdings in real time, while a Tenderly-forked Ethereum environment supports safe testing of all contract interactions before mainnet deployment.

Core Technologies and Architecture A multi-agent system lies at the heart of the off-chain service, with each Agent running continuously and maintaining trust-minimized coordination. A pruned Tendermint-based blockchain (serving as a “consensus gadget”) records key events and helps Agents agree on their internal state. Each Agent follows a finite state machine that begins by confirming its authority on the multisig Safe, then moves through rounds of data pulling, LLM-based decision-making, transaction preparation, and a pause/reset phase until the next cycle.

Security is further enhanced by requiring multiple Agent sign-offs for any portfolio action. Users can see exactly what happened at each step, including how much was traded and why, thanks to IPFS reports. Over time, the system can evolve by integrating newer LLMs or adding Agents with alternative trading models, thereby offering a growing library of AI-driven investment strategies. Trades on Uniswap V3 ensure access to deep liquidity and efficient pricing, while the multisig arrangement prevents malicious or unilateral moves of user funds.

Conclusion This project unites a decentralized off-chain consensus mechanism, LLM-based intelligence, a user-friendly web marketplace, and secure on-chain contracts to deliver an innovative DeFi portfolio management solution. By weaving together Open Autonomy, Uniswap V3, Tenderly testing environments, and IPFS for transparent reporting, it achieves a balance of automation, security, and user empowerment. Investors retain ultimate control of their assets, while benefiting from AI-generated trade insights and transparent logs of every action. Through this blend of technology and design, the project exemplifies a trustworthy and forward-looking model for autonomous portfolio optimization.