Pick the right install path
There is no single "best" way to run OpenClaw. There are four, each with a clear use case. Picking the right one now will save you a re-install later.
1. Local install
OpenClaw runs directly on your laptop or workstation.
- Best for: daily-driver assistant, full filesystem access, fastest feedback
- Trade-off: shares your machine — bugs in a Skill can read or write anywhere your user can
- Recommended for: developers, power users, people who already trust their own machine
2. Docker sandbox
OpenClaw runs inside a Docker container on your machine.
- Best for: testing untrusted Skills, isolating the agent from your real filesystem
- Trade-off: a little more friction (volume mounts), no native voice/screenshot capture
- Recommended for: experimenting with community Skills, security-cautious users
3. Separate device
OpenClaw runs on a spare laptop, mini-PC or Raspberry Pi on your local network.
- Best for: physical isolation, always-on home automation, low-cost 24/7 worker
- Trade-off: you need extra hardware, network setup is a one-time cost
- Recommended for: home-lab fans, people with a spare machine, hardware tinkerers
4. VPS
OpenClaw runs on a remote Linux server you rent (Hetzner, Hostinger, DigitalOcean, AWS Lightsail).
- Best for: 24/7 worker that survives your laptop sleeping, public-internet workflows
- Trade-off: real DevOps responsibility — security, backups, monitoring (we cover all of it)
- Recommended for: solopreneurs, anyone running an "always available" agent
How to choose for this course
You will set up #1 (local) in chapter 2 and #4 (VPS) in chapter 8. We touch #2 and #3 along the way. By the end you will have run all four and will know exactly which one fits each future project.