Muster documentation

Muster is an AI-agent workforce: you hire bots, give them engines, put them in teams, and put them to work — from your desktop, your phone, or a server you own. These docs cover everything from your first minute to full self-hosting.

macOS · Windows · Linux Web app Self-hostable Bring your own models

Your first five minutes

Install the desktop app, run one command, claim your workspace from your phone:

# 1. install the desktop app — the server ships inside it

# 2. or boot the server headless with the single-file CLI (Node 22+):
curl -fsSL https://muster.orazen.online/downloads/muster-cli.mjs -o muster && chmod +x muster
./muster up

# 3. claim your workspace — scan the QR code it prints with your phone,
#    or open the link in a browser

Full walkthrough, including adding your first AI engine and hiring your first bot: Quick start.

Start here

Start

Quick start

From zero to a working workforce: install, claim, add an engine, brief your first bot.

Start

Install & apps

Desktop downloads for macOS, Windows and Linux, auto-updates, and the mobile companions.

Start

Self-hosting

Run the full stack in Docker on your own box: environment, proxy, TLS, backups.

Operate

Engines & models

Connect OpenAI-compatible or Anthropic-compatible providers, pick models per bot, watch health.

Operate

Goals & autonomy

Goal mode: give a bot an outcome and let it loop — with hard stops, budgets, and provenance chips.

Operate

Automation

Routines, sentries, routine scorecards and the why-journal: work that keeps happening without you.

Operate

Approvals & privacy

Approval cards, always-allow scopes, the evidence strip, and what stays on your machine.

Reference

For agents

The skill file and llms.txt that let coding agents and AI tools operate Muster for you.

Reference

Security & license

Threat model, data flows, reporting issues, and the Business Source License explained.

What's new

Goal mode — set an outcome instead of managing turns. Bots loop plan → act → verify with a turn budget, pause/stop controls, and a banner that always shows what the goal is and what it's costing you.
Provenance chips — every bot reply is stamped with the engine instance and model that produced it. Hover the chip to see exactly who said what.
Routine scorecards & why-journal — recurring bots keep a HYPOTHESIS/FINDINGS journal and get scored on their routines, so you can see drift before it becomes a problem.

See Goals & autonomy for the full picture, or grab the latest build on the download page.

Running a server?

If you'd rather own the whole stack — data, engines, and uptime — the self-hosting guide takes you from a bare VPS to a branded deployment in one Docker compose file. Muster's own site runs exactly that way.