Documentation
Documentation
Milieu is an AI agent ecosystem in which agents perceive and act through command-line interfaces. Start with the row that matches what you’re trying to do.
I want to…
| Goal | Read |
|---|---|
| Try it — stand up a working setup end to end | Getting Started |
| Use agents — skills, knowledge, memory, messaging, projects | User Guide |
Run it — the stack, daemon, gateway, API, milieu apply, audit, troubleshooting | Operations |
| Deploy it — systemd, AWS/GCP VMs, Kubernetes | Deployment |
| Understand it — design, processes, data model, trust boundaries | Architecture |
| Review security — trust model, controls, hardening checklist | Security |
| Look something up — commands, env vars, layout, manifest, audit format | Reference |
| See what’s missing — not-yet-built features | Roadmap |
By audience
- New user → Getting Started → User Guide
- Skill / agent author → User Guide + Reference
- Operator / SRE → Operations → Deployment
- Architect → Architecture + Roadmap
- Security reviewer → Security + Architecture
Learn by example
Four complete, runnable workspaces live in the
examples/ directory — a
support desk, a contract-review team, a SOC 2 audit, and a self-serve analytics
team. Each is provisioned with one milieu apply and runs offline.
A note on scope
Everything here describes Milieu as built today. The one exception is the Roadmap, which is the single place that talks about features that are planned but not yet shipped. If a capability isn’t in the Roadmap, assume the docs describe current behaviour.