> ## Documentation Index
> Fetch the complete documentation index at: https://docs.zhentan.me/llms.txt
> Use this file to discover all available pages before exploring further.

# Overview

> Technical overview of how Zhentan is built.

Zhentan is built on three independently runnable components that work together: an instant deterministic scoring layer, an interactive AI agent layer, and a Safe multisig execution layer where **the user always holds the majority of keys**.

<CardGroup cols="2">
  <Card title="Signing & Wallet Profiles" icon="signature" iconType="duotone" href="/technology/signing">
    The 2-of-3 model — profiles, transitions, relay-only execution, and the Safe Transaction Service flow.
  </Card>

  <Card title="Architecture" icon="diagram-project" iconType="duotone" href="/technology/architecture">
    System overview, component breakdown, and data flow diagrams.
  </Card>

  <Card title="Risk Scoring" icon="gauge-high" iconType="duotone" href="/technology/risk-scoring">
    How transactions are scored 0–100 against behavioral patterns.
  </Card>

  <Card title="Agent" icon="robot" iconType="duotone" href="/technology/agent">
    The NanoBot/Hermes skill pack — skills, lifecycle, and Telegram commands.
  </Card>

  <Card title="Legacy Accounts" icon="clock-rotate-left" iconType="duotone" href="/technology/legacy">
    How pre-refactor 2-of-2 accounts work and how they upgrade to the full model.
  </Card>

  <Card title="Contract Deployments" icon="file-contract" iconType="duotone" href="/technology/contracts">
    Safe and infrastructure contract addresses on BNB Chain.
  </Card>
</CardGroup>

## Design Principles

* **The user holds the majority** — in the full 2-of-3 model your two keys meet the threshold. The agent is an advisory speed bump you can always go around via the Safe app, never a gatekeeper.
* **The agent never signs what it didn't screen** — when your signatures alone meet the threshold, the agent relays and pays gas without contributing a signature. Enforced server-side at proposal and again at execution.
* **No LLM on the hot path** — inline risk scoring is deterministic and instant. The agent only steps in after a verdict is reached.
* **Safe multisig is the foundation** — every transaction is a standard SafeTx, visible and actionable in the official Safe app at every stage. Zhentan adds intelligence on top without replacing key management.
* **Continuous learning** — patterns are recorded after every confirmed transaction. The profile improves automatically.