Getting Started¶

Missy is a security-first, self-hosted AI assistant for Linux. It runs entirely on your own hardware, enforces strict access policies on every operation, and provides a full audit trail of everything it does.

Who is Missy for?¶

• Developers and sysadmins who want an AI assistant that respects security boundaries
• Teams that need auditability and policy enforcement for AI-assisted workflows
• Privacy-conscious users who prefer self-hosted over cloud-managed agents
• Homelab enthusiasts interested in voice-controlled AI with Raspberry Pi edge nodes

What you will learn¶

This guide walks you through going from zero to a working Missy installation:

How Missy works¶

Every capability -- network access, filesystem writes, shell commands, plugins -- is disabled by default. You explicitly enable what you need in ~/.missy/config.yaml. The policy engine enforces these rules on every request, and every action is logged to a structured audit trail.

CLI / Discord / Webhook / Voice
         |
    AgentRuntime
    +-- PolicyEngine (network / filesystem / shell)
    +-- InputSanitizer + SecretsDetector
    +-- ProviderRegistry (Anthropic / OpenAI / Ollama)
    +-- ToolRegistry + MCP servers
    +-- Memory + Learnings
    +-- AuditLogger

Missy supports three AI providers out of the box:

• Anthropic (Claude) -- recommended default
• OpenAI (GPT-4o, Codex)
• Ollama (local models, fully offline)