The strongest reason to use OpenClaw is also the strongest reason to self-host it: control. The official docs position OpenClaw as running on your own hardware or server, with config, sessions, workspace, and channel routing under your control.
The basic self-hosted flow
- install the CLI
- run onboarding
- choose local or remote gateway mode
- configure channels and providers
- open the dashboard and test before expanding
What the docs currently show
The official quick start uses:
npm install -g openclaw@latest
openclaw onboard --install-daemon
openclaw dashboardWhy self-hosting matters
- more control over data locality
- more control over channels and credentials
- more control over skills, plugins, and policy
What self-hosting does not magically solve
It does not remove risk. You still need secure config, tighter file permissions, careful plugin choices, and good channel policies.
Good self-hosting habits
- limit who can message the agent
- tighten config file permissions
- review plugins and skills before enabling them
- keep backups of config and state
Useful next reads
Read OpenClaw security risks: what developers should know before automating everything and OpenClaw skills and plugins: which ones are actually useful today.
Quick FAQ
Can I run OpenClaw on a VPS?
Yes. The docs explicitly support local or remote gateway patterns.
Where is the config stored?
The docs point to ~/.openclaw/openclaw.json as the main config location.