one-command.md

One command — the cloud installer

This is Path 1: you have a fresh Linux VM or bare-metal host and an SSH session. One line turns it into a DarkMoon node.

curl -fsSL https://portal.dark-moon.org/cloud | sudo DARKMOON_LICENSE_KEY=<KEY> bash -s -- \
  --provider anthropic --model claude-opus-4-6 --api-key sk-ant-...

What it does

The URL serves deploy/cloud-install.sh. Piped into sudo bash, it runs three steps in order:

1. Installs Docker — Docker Engine + Compose v2, via the official get.docker.com convenience script.
2. Runs install.sh non-interactively — it fetches the DarkMoon installer and runs it with the flags you passed after --. install.sh requires the Pro license (via DARKMOON_LICENSE_KEY or as the first positional argument) and a provider selection.
3. Runs darkmoon doctor — verifies runtime, containers, images, license, AI provider, Privacy Gateway, ports, disk and memory before handing back.

Everything after bash -s -- is passed straight through to install.sh.

The license (required)

DarkMoon Pro will not install without a valid license. Provide it either as an environment variable (recommended, shown above) or as the first positional argument to install.sh:

# env var form (preferred)
sudo DARKMOON_LICENSE_KEY=<KEY> bash -s -- --provider anthropic --model claude-opus-4-6 --api-key sk-ant-...

# first-positional form
sudo bash -s -- <KEY> --provider anthropic --model claude-opus-4-6 --api-key sk-ant-...

Each fresh VM consumes a device/machine slot on the license (machine-code fingerprint). See Doctor & lifecycle before you script the creation of many ephemeral VMs.

Provider flags

Pick exactly one AI backend. See AI modes for what each means.

Connected (a hosted LLM provider)

--provider <name>      # e.g. anthropic
--model   <model-id>   # e.g. claude-opus-4-6
--api-key <key>        # e.g. sk-ant-...

A native Anthropic endpoint (explicit URL/model/key)

--anthropic-url   <base-url>
--anthropic-model <model-id>
--anthropic-key   <key>

Use the native anthropic path (not an OpenAI-compatible shim) when you want Anthropic prompt caching honoured.

Local (on-node inference, no LLM egress)

--local
--local-engine ollama|llama.cpp
--local-url    <endpoint>       # e.g. http://127.0.0.1:11434
--local-model  <model-id>

Local mode needs extra RAM/VRAM per model — see Sizing and AI modes.

Common combinations

# Connected — hosted provider
curl -fsSL https://portal.dark-moon.org/cloud | sudo DARKMOON_LICENSE_KEY=<KEY> bash -s -- \
  --provider anthropic --model claude-opus-4-6 --api-key sk-ant-...

# Private — customer LLM endpoint via the native anthropic path
curl -fsSL https://portal.dark-moon.org/cloud | sudo DARKMOON_LICENSE_KEY=<KEY> bash -s -- \
  --anthropic-url https://llm.internal.example.com --anthropic-model claude-opus-4-6 --anthropic-key <KEY>

# Local — on-node inference, no LLM egress
curl -fsSL https://portal.dark-moon.org/cloud | sudo DARKMOON_LICENSE_KEY=<KEY> bash -s -- \
  --local --local-engine ollama --local-url http://127.0.0.1:11434 --local-model <model-id>

After it finishes

• Open the dashboard on the node's ui_port (default 80). Restrict that port to trusted CIDRs — see Security.
• Re-run diagnostics any time with darkmoon doctor.
• Operate the node through the darkmoon CLI, not the raw containers — see Doctor & lifecycle.

Security reminder for automation

When you script this (user-data, Terraform, CI), do not inline the license key or the API key in plaintext where instance metadata or the console can read them. Fetch them from the cloud secret manager at boot. See Security.

Review before you pipe

Piping a URL into sudo bash executes remote code as root. If your policy requires review first, download and read it, then run it:

curl -fsSL https://portal.dark-moon.org/cloud -o cloud-install.sh
less cloud-install.sh
sudo DARKMOON_LICENSE_KEY=<KEY> bash cloud-install.sh --provider anthropic --model claude-opus-4-6 --api-key sk-ant-...