---
title: "Security Overview"
description: "Nexus OS is designed with security as a core principle. Every component follows the principle of least privilege. Nexus OS uses a layered security model: WASM Sandboxing — Agents run in isolated WebAssembly sandboxes wit"
resource: https://www.aiagents.nexus/docs/manual/security/overview
generated: { by: "process:nexus-agent-assets", at: 2026-09-07T09:13:03Z }
status: stable
---

# Security Overview

Nexus OS is designed with security as a core principle. Every component follows the principle of least privilege.

## Security Model

Nexus OS uses a layered security model:

```
┌─────────────────────────────────┐
│  Application Layer              │
│  Agent code + business logic    │
├─────────────────────────────────┤
│  Orchestration Layer            │
│  Supervisors, Sagas, Workflows  │
├─────────────────────────────────┤
│  Sandbox Layer                  │
│  WASM isolation + syscall filter│
├─────────────────────────────────┤
│  Storage Layer                  │
│  Encrypted SQLite + audit trail │
└─────────────────────────────────┘
```

## Key Security Features

- **WASM Sandboxing** — Agents run in isolated WebAssembly sandboxes with no direct filesystem or network access
- **AXIS Trust Verification** — Cryptographic verification of agent identity and capabilities
- **Audit Trail** — Every action is logged with timestamps, agent IDs, and details
- **Cost Controls** — Budget enforcement prevents runaway spending
- **Credential Isolation** — API tokens and secrets are never exposed to agent code

## Threat Model

| Threat | Mitigation |
|--------|------------|
| Malicious agent code | WASM sandbox prevents escape |
| Credential theft | Secrets injected at runtime, never stored in agent code |
| Cost attacks | Per-agent budgets with hard limits |
| Data exfiltration | Network access controlled by sandbox policy |
| Privilege escalation | No root access, minimal syscall surface |