Architecture

More details coming soon.

At its core, Armor Wallet leverages a multi-agent system (MAS) architecture, where user requests are parsed by a natural language chat interface and delegated to specialized agents for execution. These agents handle everything from token swaps to portfolio management, enabling users to perform complex trades using plain English (or any other language). This system is further supported by a planner agent, which coordinates tasks and offloads delayed or conditional actions to background workers. By abstracting away technical complexity, Armor Wallet empowers both new and experienced users to engage confidently with DeFi and crypto markets.

Underpinning this intelligent agent system is a robust technical foundation that combines real-time blockchain data ingestion, advanced transaction simulation, and layered security protocols. Armor Wallet integrates high-speed ETL12 pipelines and a custom-built data mesh that aggregates on-chain and off-chain signals, including market activity, social sentiment, and token metadata, into a unified knowledge base accessible to AI agents. Each transaction is simulated for safety, checked against blacklists, and subject to user-defined 2FA controls, ensuring secure execution. With its unique WYAIWYG (What You Ask Is What You Get) approach, Armors' AI can even generate new tools and features on demand, giving users unprecedented power, flexibility, and control in a single, intuitive platform.

Real-Time Interaction

In current AI systems, the AI is passive. It waits for the user to engage before it does anything. Armor is different in that the Armor AI will engage with a user proactively. Armor Agents can research, perform transactions, swaps, and other functions while completing tasks for users. Imagine an agent doing some research into a couple of markets and the blockchain. While this research is going on, users can continue their conversation, and the agent will notify them when it has an answer for the research. This makes Armor Agents a much more meaningful tool to help users invest better and stay on top of their finances in real time.

WYAIWYG

A fundamental shift in how Armor is approaching the UX is WYAIWYG, What You Ask Is What You Get. In the use of Armor Wallet, the AI not only accesses functions and data but also writes code to create new features and strategies that can be utilized and monetized inside the Armor Ecosystem. This gives Armor an infinite possibility of features, behaviors, and actions.

A new trend that has recently become popular is "vibe coding" where users prompt an LLM for code, embrace exponentials, and just go with it (with testing and feedback, of course). Armor implements the WYAIWYG methodology in the Armor Agent architecture. Think of it as "vibe trading". When users ask an agent to perform a complex task, it will think about the optimal solution and code it for them. The code output is run through a code sanitizer, and then the agent will explain and ask to verify what is about to happen. On a user's mark, the Armor Agent will run the code and perform swaps, manage wallets, or whatever they ask that is within the set of available features.

Context & Recall

Graph Retrieval-Augmented Generation14 (GraphRAG) enhances Armor Wallet’s Multi-Agent System (AMAS) by providing a more contextually rich and logically structured information retrieval mechanism. Unlike traditional Retrieval Augmented Generation15 (RAG), which relies solely on semantic similarity, GraphRAG integrates knowledge graphs to maintain relationships between logical rules, financial taxonomies, and transaction dependencies. In Armor Wallet, where users request complex, multi-step financial operations, semantic recall alone is insufficient to retrieve the most relevant data. By leveraging GraphRAG, Armor Wallet can accurately interpret user intent, ensuring that AI agents receive precise and logically structured insights when executing swaps, portfolio rebalancing, and multi-condition trades. This minimizes errors and enhances the accuracy of transaction planning by preserving relationships between assets, trading strategies, and execution conditions while keeping all of the data within the model context window.

The integration of GraphRAG within Armor Wallet’s tech stack significantly improves retrieval efficiency and decision-making across AI agents. GraphRAG’s ability to reduce computational overhead aligns with Armor Wallet’s goal of fast and efficient execution by eliminating the need for excessive embedding-based recall. This allows specialized agents, such as those handling swaps, liquidity management, or risk assessment, to access relevant knowledge graphs instantly, leading to faster transaction processing and improved security checks. Additionally, because GraphRAG maintains relationships between financial data points, it enables dynamic risk assessments, fraud detection, and smarter AI-driven trading strategies. By integrating GraphRAG, Armor Wallet ensures that users benefit from contextually aware, accurate, and computationally optimized AI-driven financial operations, reinforcing its position as an intelligent and secure crypto trading platform.