> For the complete documentation index, see [llms.txt](https://maiga.gitbook.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://maiga.gitbook.io/docs/maiga-tech/how-it-works.md).

# How it works?

{% hint style="success" %}
Maiga.ai redefines what it means to trade, earn, stake, and profit, in the era of Web3.0 and #DeFAI.
{% endhint %}

## How Maiga AI agent works?

<figure><img src="/files/E5tSlNlxk4U0A006Zqr9" alt=""><figcaption></figcaption></figure>

## Creation, Paths, LLMs & MPC

#### **Maiga AI agent creation:**

* AI agents are created using Maiga’s native token, $MAIGA. This token acts as the foundation for activating and owning these agents.
  * **Example:** A user might spend $MAIGA to create an agent specialized in monitoring and trading DeFi assets. Once created, the AI agent is ready to receive tasks and training for personalized use cases.
  * **Technical:** The creation process initializes a connection between the AI agent and Maiga’s shared knowledge base, equipping with default libraries and capabilities.

#### **AI Agent Paths:**

* These are pre-built guides or roadmaps that help Maiga AI agents navigate blockchain networks efficiently.
  * **Example:** If an AI agent needs to perform a cross-chain token swap, it will use a specific path that outlines the necessary steps, including locating a bridge, ensuring liquidity, and completing the transaction.
  * **Technical:** These memories are maintained as part of a graph database, where nodes represent specific blockchain operations (e.g., smart contracts, DEXs). AI agents query the graph to identify optimal routes for task execution, reducing computational costs and errors.

#### **Reinforced Learnings:**

* AI agents can learn from past interactions, improving their decision-making and task efficiency over time.
  * **Example:** An AI agent trading agent might recognize patterns in market behavior and adjust its strategies to maximize returns based on previous successes and failures.
  * **Technical:** Memory systems categorize experiences into short-term (active tasks), long-term (cumulative knowledge), and pinned memories (key learnings). This architecture ensures agents adapt and refine their actions based on evolving conditions.
  * **Advanced Training:** Users can further enhance agents by providing domain-specific data, such as financial reports or specialized tutorials. This tailored training enables agents to specialize in tasks like market analysis, sentiment analysis or more.

#### **MPC Wallet Management:**

* AI agents come equipped with digital wallets, enabling them to securely store, trade, and manage blockchain assets.
  * **Example:** An AI agent might manage an EVM wallet, autonomously transferring funds to a staking protocol while ensuring compliance with user-defined conditions.
  * **Technical:** Maiga’s platform leverages Multi-Party Computation (MPC) wallet architecture. This ensures private keys are never fully exposed, adding an abstraction layer for enhanced security.
    * **Custom Wallets:** Users can import dedicated wallets to their agents for specific purposes. For instance, a wallet can be assigned to an AI agent managing DeFi strategies while isolating other funds.
    * **Abstraction Benefits:** By abstracting wallet management, agents can interact seamlessly across blockchains without compromising security. This abstraction also allows for smoother cross-chain operations, aligning with Maiga’s interoperability goals.

<br>