Workflow

Self-host your assistant (the safe way)

From zero to an always-on OpenClaw on a Pi, an old laptop, a Mac mini, or a $5 VPS — the setup path, and the security settings that are not optional.

Tools in this workflow

Goal: an assistant that’s always on, on hardware you control — without accidentally publishing your inbox to the internet.

Pick the box — anything that stays on works: a Raspberry Pi (cheapest 24/7), the laptop in the drawer, a Mac mini (best if you want iMessage), or a small VPS. Cloud guides exist for AWS and friends; a VPS is simpler.
Run the wizard — install OpenClaw and run openclaw onboard. It walks the gateway, workspace, model keys, and first channel end to end. Prefer containers? There’s a Docker path — pin a tag, snapshot before upgrades.
Pair your channels — WhatsApp by QR scan, Telegram by bot token, and so on down the channels list. Start with one.
Do the security pass before the fun — keep the defaults that make unknown DMs get a pairing code instead of an answer; leave the gateway bound to localhost; read the security guide and the exposure runbook before opening anything to the internet; treat community skills as unaudited code you read first.
Verify — openclaw doctor, then text it “what machine are you on?” from your phone.

From here, every OpenClaw workflow on Skillra is a copy-paste away — start with The 7 a.m. briefing.