muthr

muthr is a zero-trust orchestrator for local AI on Apple Silicon. It runs inference engines on the host, isolates agent runtimes inside per-project containers, and provides stable service endpoints for MCP and search tooling.

muthr follows Unix design principles:

• subcommands grouped by lifecycle domain
• scriptable output (text, json, ndjson)
• explicit defaults you must override intentionally
• safe behavior when context is missing

Why this exists

Agent stacks execute package installers, shell commands, generated scripts, and network clients with broad filesystem access. A compromised dependency gives an attacker access to your credentials, keys, and system files.

muthr reduces that risk through enforced boundaries:

• run agent execution inside sandbox containers
• keep host-only assets outside container mounts
• gate runtime context with explicit env contracts
• log every sandbox session for audit and incident review

Core model

1. Host inference runtime
   • mlxcel-server, llama-server, or vllm-mlx managed by muthr engine
2. Persistent services plane
   • muthr-services + muthr-searxng managed by muthr services
3. Per-project sandbox
   • one container per project directory, managed by muthr sandbox

The default muthr run path boots engine + services, then you enter project sandboxes as needed.

Read next

1. Quickstart
2. Commands
3. Sandbox Architecture
4. Security

Acknowledgements

muthr depends on these projects:

• llama.cpp
• mlxcel
• vllm-mlx
• Apple container

---

See also: Quickstart · Concepts · Security