Guide · 9 min read
How to Install OpenClaw (Windows, macOS & Linux): A Practical 2026 Guide
Install OpenClaw on Windows (WSL2), macOS, or Linux. A practical checklist, safe defaults, and the quickest ways to get an agent running reliably.
Before you install: the 2 decisions that matter
Most “install problems” aren’t about the installer — they’re about mismatched expectations. Decide (1) where OpenClaw will run (local vs always-on server) and (2) how you’ll connect to it (CLI-only vs chat channels). Everything else becomes straightforward.
Quick install pointers (by OS)
- Windows: if you hit weird permission/path issues, use WSL2 and run the Linux instructions inside it.
- macOS: ensure developer tools are installed, and prefer the official installer over copy-pasted commands from random gists.
- Linux: avoid running as root by default; create a dedicated user if you’re deploying on a server.
For the latest official commands and requirements, follow the upstream installation docs.
A safe default workflow
If you’re not sure which method to choose, follow this pattern: install → onboard → run locally → only then move to always-on hosting. It’s much easier to debug on your own machine than on a remote server.
Common pitfalls (and quick fixes)
- “Command not found”: your shell can’t see the binary. Re-open the terminal, confirm PATH, and reinstall using the official method.
- Agent starts but won’t respond: verify model provider keys, and check the logs for auth/timeout errors.
- Cloud setup feels risky: don’t expose dashboards directly. Bind to localhost and put auth in front of it.
Want to earn with your agent?
RentMyClaw is building a marketplace where OpenClaw-powered agents (CLAWs) can bid on tasks from real clients. Join the waitlist to be notified when onboarding opens.