Question 1

What is OpenClaw?

Accepted Answer

OpenClaw is an open-source AI assistant that runs on your own devices or server. It connects models, tools, messaging channels and optional companion apps through one Gateway. It is not a rename of any other project.

Question 2

How much RAM does OpenClaw need?

Accepted Answer

There is no single published runtime minimum. The official Docker guide lists at least 6 GB RAM for building the image from source locally, and notes that pre-built images avoid that build requirement. Size the rest around the tools, channels and any model server you add.

Question 3

Does OpenClaw need a GPU?

Accepted Answer

No. A GPU only matters if you separately run a local model on the same host. If you connect hosted model providers, a CPU-only server is enough.

Question 4

Is Docker required for OpenClaw?

Accepted Answer

No. The official docs describe Docker as optional and useful for an isolated gateway environment. The standard path is the maintained installer described at docs.openclaw.ai/install.

Question 5

Can I use shared hosting for OpenClaw?

Accepted Answer

Shared hosting normally gives you no root access and no long-running process control, so a VPS or comparable server is the practical option.

Question 6

How do I keep a public VPS deployment safe?

Accepted Answer

Do not publish the Gateway port directly to the internet. Bind it locally and reach it through a VPN, SSH tunnel or authenticated reverse proxy, and follow the official security hardening notes referenced in the Docker install guide.