Architecture

Maiga AI technology architecture framework

Architectural overview

Maiga AI’s architecture is designed to be a chain-agnostic, scalable foundation enabling AI agents to interact seamlessly across multiple blockchain ecosystems.

• Core Knowledge Base: At its heart, Maiga AI agent provides a robust repository of libraries for paths and tools that enable AI agents to manage wallets, execute transactions, and interact with decentralized applications (dApps). This knowledge base includes critical details such as blockchain addresses, fees, and operational protocols.

  • Example: If an AI agent needs to execute a token swap, it can quickly access the required path and complete the task efficiently without requiring repetitive data retrieval.

• Blockchain Ecosystem Graph: Upon launch, Maiga will include a dynamic graph of major blockchain entities and smart contracts. This graph maps relationships within and between ecosystems, allowing AI agents to autonomously traverse these networks.

  • Technical Insight: The graph integrates metadata about smart contracts, transaction histories, and user preferences, enabling AI agents to make informed decisions.

• Private Key Management: A vital component of the MPC wallet architecture ensures secure handling of private keys for the wallets controlled by AI agents. These keys are encrypted and tied to specific shells, ensuring tamper-proof operations.

  • Example: An AI agent can use its private key to sign transactions without exposing sensitive information, maintaining the highest levels of security.

Maiga AI’s technical approach

Maiga AI is engineered as a high-performance, omni-chain tool, initially residing on the EVM blockchain but optimized for compatibility with other blockchain ecosystems. This chain-agnostic design ensures:

1. High-Speed Transactions: Leveraging EVM’s efficiency, Maiga AI can handle time-sensitive tasks, such as executing trades in volatile markets.

2. User-Friendly Interactions: Simplified interfaces allow users to manage complex blockchain operations through intuitive commands and prompts.

3. Advanced Memory Retention: Maiga AI maintain detailed transaction logs and learning histories, enabling iterative improvements in decision-making.

Consensus and verification

Maiga AI employs a dual-layer verification system to ensure the reliability and consistency:

1. Verification Agents: Trusted community-maintained AI agents automatically evaluate the safety and accuracy of insights, signals and execution.

2. Manual Oversight: Human validators review complex or high-stakes paths, adding an additional layer of assurance.

   • Example: If a developer submits a new path for cross-chain swaps, it undergoes automated checks and human review before being added to the shared library.

Purpose-built LLMs

Maiga AI integrates customized large language models (LLMs) to enhance shell functionality:

1. Specialized Models: Compact LLMs tailored for blockchain operations, such as EVM function calls, ensure efficiency and accuracy.

   • Example: A fine-tuned LLM can parse complex user commands, like "Convert 10% of my USDT to WETH if gas fees are below $2," and execute them seamlessly.

2. Open-Source Philosophy: Both the datasets and model weights are open to the community, fostering transparency and collaborative innovation.

User-friendly interface for non-coders

Maiga AI prioritizes accessibility, providing tools that simplify shell creation and management:

1. User Interface (UI): Users can design AI agents, set operational parameters, and issue commands through an intuitive chat platform.

   • Example: A user creates a AI agent that monitors news for major Web3 announcements and executes pre-set trades based on keywords.

2. Comprehensive Monitoring: Dedicated dashboards allow users to view AI agent activities, manage memory, and update configurations in real-time.

Comprehensive integrations for developers

Developers benefit from seamless integrations with popular frameworks like Eliza OS and Automate by HeyAnonAI.

• Enable access to composable DeFAI datasets and blockchain tools.

• Provide API-based interactions for custom AI agent development.

AI agent operations and "Explicit Intent"

While Maiga AI agent are useful, they operate within clearly defined boundaries to ensure user control:

• Bounded Autonomy: AI agent execute tasks only within the parameters set by their owners. For example, trading on a DEX will adhere strictly to the user-defined slippage and budget limits.

• Confirmation Mechanisms: Before executing critical transactions, AI agent seek user approval, ensuring transparency and control.

  • Example: An AI agent directed to stake tokens in a new protocol prompts the owner for confirmation, detailing potential risks and rewards.