Hunch

Hunch is a decentralized application (dApp) designed to dismantle the walled garden of professional crypto trading. It creates a trustless, non-custodial marketplace where expert strategy creators can securely monetize their trading intelligence, and everyday investors can access advanced, automated strategies entirely through a simplified, conversational chat interface. The core innovation of Hunch lies in solving the intellectual property (IP) risk for creators while ensuring self-custody for users, all while making the entire platform user-friendly.

The entire user experience, for both creators and investors, is mediated by a conversational AI agent. This chat-first interface eliminates complex dashboards, managing all tasks—from risk assessment and strategy discovery to trade execution and profit withdrawal—through natural language. The agent uses a knowledge graph to provide personalized recommendations and dynamically generates charts and UI elements directly within the chat window, making sophisticated trading accessible to everyone.

The platform functions as a decentralized agent marketplace for two types of assets: high-level strategies that automate trading, and lower-level indicators that provide custom metrics. A core feature is multi-layer monetization, allowing creators to build advanced strategies by securely nesting and licensing other creators' proprietary indicators, which guarantees royalty streams for underlying IP without revealing the indicator's secret logic. All agents are ranked by an immutable, on-chain reputation score derived from community governance, fostering trust and transparency across the marketplace.

The project's technical innovation lies in the trustless & private execution engine. To preserve the creator's intellectual property ("alpha"), the trading logic is client-side encrypted and then processed by decentralized executor nodes. Critically, users maintain non-custody over their funds; they simply grant a limited, revocable trading allowance to a secure executor smart contract. This contract authorizes the cryptographically proven trade instruction on a decentralized exchange. Upon successful trade, an automated profit-sharing smart contract transparently distributes the agreed-upon percentage of gains to the user, the strategy creator, and any nested indicator creators, completing the end-to-end decentralized commerce cycle.

Hunch is built on a four-tiered architecture designed for decentralization, intelligence, and a delightful user experience. The core challenge was integrating the conversational AI brain with the non-custodial execution layer while protecting intellectual property. The entire application is centered on the AI Agent, powered by the ASI (Fetch.ai) uAgents Framework and the MeTTa Knowledge Graph (KG), which provides the critical ability to reason, converse, and automate complex financial logic. All core logic—from risk assessment to transaction construction—is defined as skills within the Python uAgent, which interprets natural language. The MeTTa KG is pre-loaded with complex financial and risk-based rules, enabling the AI to make intelligent, rule-based recommendations that go far beyond simple data retrieval. It also leverages the Knowledge Graph for risk assessment and user behavior modeling, ensuring adaptive personalization across strategies.

We chose Hedera as the settlement and execution layer due to its finality and low transaction costs. We deployed two main Solidity Smart Contracts on the Hedera EVM: the Subscription/Governance Contract, which also enables pub-sub messaging and community voting/review of strategies, and the Profit-Sharing Contract for automatic multi-party payout.

For lightweight storage, the platform uses SQLite to persist non-critical user metadata such as chat history, allowing the frontend to retrieve information quickly without compromising decentralization of core financial logic.

Finally, the user interface is a Next.js application. Our most notable "hack" is the Generative UI: instead of sending back simple text, the uAgent returns a structured JSON object containing a component type and relevant data. The Next.js frontend is built with a library of pre-made React components and dynamically renders the appropriate visual component (like a performance chart or a "Confirm Trade" button) based on the JSON, keeping all sophisticated interaction entirely within the conversational flow.