How to Install OpenClaw on Windows (WSL2 Guide)

The short version of installing OpenClaw on Windows: do it inside WSL2, not native PowerShell. OpenClaw is built around a Unix-style environment — shell commands, file paths, a background daemon — and WSL2 gives it exactly that, running inside Windows with almost no performance cost. People who fight to run it directly on Windows usually end up switching to WSL anyway, so we’ll start there and save you the detour.

This guide covers the prerequisites, getting WSL2 ready, installing OpenClaw, connecting a model, and the gotchas that trip people up. If you’re brand new to OpenClaw, the explainer gives the background; otherwise, open a terminal.

Before you start#

If WSL2 isn’t set up yet, our WSL2 install guide gets you there, and Microsoft’s WSL environment setup covers installing Node and git inside the distro. Come back once wsl drops you into a Linux shell and node -v works.

Step 1: open your WSL shell#

Everything from here happens inside WSL, not PowerShell. Launch your distro:

wsl

You’re now in Linux. Confirm the basics are present:

node -v
git --version

If either is missing, install it inside the distro before continuing — Node and git living in Windows don’t count here; OpenClaw needs them in the Linux environment.

Step 2: get OpenClaw#

Clone the repository into your WSL home directory and install dependencies. The exact install steps can change as the project moves, so cross-check the OpenClaw repo README, but the shape is:

cd ~
git clone https://github.com/openclaw/openclaw.git
cd openclaw
npm install

Some releases offer a one-line install script instead of a manual clone. If the README points to one, prefer it — it handles the steps above and any extras the current version needs.

Step 3: connect a model#

OpenClaw needs a brain before it’s useful. You set the provider’s endpoint and key, then choose the model. Keep the key in an environment variable rather than pasting it into a config file:

export OPENCLAW_API_KEY="your-provider-key"

Which provider you point at is the decision that controls cost:

For the lowest-friction cheap start, follow the DeepSeek setup; for fully free and private, the Ollama local setup. Both pick up exactly where this guide leaves off.

Step 4: run the daemon#

Start OpenClaw. Again, confirm the current command against the repo, but it’s typically a start script:

npm start

On first run it walks you through configuration — picking the model, choosing a channel to talk through, enabling skills. Start by talking to it in the terminal; you can add Telegram or Discord later once it works.

Step 5: confirm it works#

Talk to it. Ask it something simple that exercises a capability — “what files are in this directory?” — and check it responds and acts. If the model answers but can’t touch files, the model is connected but a permission or skill isn’t enabled; if nothing answers, the model connection is the issue.

Common gotchas#

A few things that cause most first-install trouble:

• Running in PowerShell instead of WSL. The number-one source of pain. If setup is fighting you on native Windows, move to WSL2 and most of it evaporates.
• Node/git installed on Windows but not in the distro. They have to exist inside WSL. Check with node -v in the WSL shell.
• Working from /mnt/c. Slow and permission-prone. Use your Linux home.
• Key in a config file that gets shared. Keep it in an environment variable so it doesn’t leak.

Wrapping up#

Installing OpenClaw on Windows is mostly about choosing the right environment: WSL2, not native PowerShell. Get WSL2 and a distro ready with Node and git, clone OpenClaw into your Linux home, connect a model with the key in an environment variable, and start the daemon. Talk to it in the terminal first, then expand to messaging channels and skills once it’s working.

Next, give it a brain on a budget with the DeepSeek setup or go fully local with Ollama. The power-user guide maps everything else you can do from here.