The problem Horus solves

DeFi users often have complex, multi-layered positions across various protocols. In the event of a protocol exploit, stablecoin depegging, or systemic failure, users may not realize their exposure until it’s too late.

Hidden Risk Exposure – Funds can be indirectly at risk even if users are multiple levels removed from an affected asset.
Slow Reaction Time – Manually identifying exposure and executing transactions across multiple platforms is time-consuming.
Event-Driven Response – Most risk monitoring tools focus on historical data or vague risk predictions, rather than immediate action.
This AI agent reacts instantly by detecting exposure, identifying the safest exit strategy, and executing fund withdrawals automatically, protecting users before they suffer losses.

User Interaction and Data Flow

1. User Connects Wallet – The AI agent scans all on-chain positions, identifying assets and protocol interactions.
2. AI Monitors Risks in Real-Time – Continuously watches for major security events (hacks, depegs, insolvencies) using news feeds, on-chain activity, and risk reports.
3. Risk Exposure Mapping – The AI recursively traces exposure across multiple layers, understanding direct and indirect risks.
4. Immediate Action – Upon detecting risk, the AI agent automatically withdraws or rebalances assets to a safer position based on predefined user settings. This can include swaps, revokes, withdrawals, and many more.
5. User Notifications – The user receives instant alerts detailing the event, exposure level, and executed actions.

The project architecture and development process

Solution Overview

Our AI-powered Wallet Security Agent detects and reacts to DeFi threats like stablecoin depegs and protocol hacks. It monitors on-chain and off-chain security signals, evaluates user exposure, and executes protective actions to mitigate risk.

System Architecture

The agent follows a threat detection and response pipeline:

1. Threat Signals → Monitors security events (e.g., depegs, hacks).
2. Signal Processor → Analyzes what’s happening in the ecosystem.
3. Evaluation Module → Checks if the threat affects user assets (recursive risk mapping).
4. Threat Evaluation Action Composer → Generates contract calls for automated risk mitigation.
5. Automated Execution → Executes withdrawals, swaps, revokes, or reroutes using smart contract automation.

Development Process

We followed an incremental build approach, focusing on an MVP first and adding more automation over time.

1. MVP Phase → Integrated threat detection, simple automated withdrawals, and basic swaps.
2. Expansion Phase → Added recursive risk evaluation, revocation actions, and EigenLayer AVS integration.
3. Optimization Phase → Improved gas efficiency, custom risk preferences, and real-time dashboards.

Key Takeaways

• Speed matters → Built a low-latency event processor for instant security response.
• DeFi is complex → Implemented multi-layered risk tracing.
• User control vs. automation → Added manual overrides for flexibility.

Product Integrations

Chainlink, CMC, Uniswap – Track real-time price fluctuations and stablecoin depegging.
Twitter – For news feeds
AgentKit – For AI and wallet actions
AVS – For proving off chain execution on chain

Key differentiators and uniqueness of the project

Instant Automated Protection – Unlike traditional risk monitoring tools that only notify users, this agent actively executes protection strategies without requiring manual intervention.
Multi-Layered Risk Mapping – Tracks exposure across deeply nested positions, revealing risks that are otherwise difficult to detect.
Event-Driven Response – Instead of trying to predict risks, it reacts instantly to real-world events (hacks, depegs, insolvencies).
Gas-Efficient Exit Strategies – Optimized transaction execution to minimize slippage and losses when withdrawing funds.
Configurable User Preferences – Users can set risk thresholds and custom response strategies (e.g., move assets to cold storage, convert to stablecoins).

Trade-offs and shortcuts while building

Initial Centralized Data Sources – While the goal is full decentralization, the first iteration relies on centralized news feeds and APIs for real-time event tracking. Future versions will integrate decentralized monitoring solutions.
Protocol-Specific Exits – Some DeFi protocols have complex withdrawal mechanisms, requiring protocol-specific strategies. The AI currently supports major platforms, with more integrations planned. These integrations are done manually
Twitter Feed Input – Manually scraped at the moment, but will be automated by AI to retrieve from reputable sources