How to install OpenClaw on Windows 11

OpenClaw is a self-hosted AI agent that runs locally on Windows 11 or in the cloud.
Unlike ChatGPT or Claude, it automates tasks instead of requiring constant prompts.
It connects to AI providers like OpenAI and Anthropic using API keys.
Skills and jobs allow continuous background automation across apps like Gmail, Slack, and GitHub.
Installation is available via PowerShell or WSL with optional onboarding.

On Windows 11, you can install and configure OpenClaw to automate repetitive tasks with AI, reducing manual effort and helping you stay focused on higher-value work.

What is OpenClaw?

OpenClaw is a self-hosted AI assistant (recently acquired by OpenAI but still open-source on GitHub) that runs locally on your device or in a cloud environment, such as a VPS. Unlike traditional chatbot experiences such as ChatGPT or Claude, it functions as an autonomous agent rather than a tool you interact with on demand.

To operate, it must be connected to an AI provider such as OpenAI or Anthropic via API keys, which typically require a subscription. This connection enables the assistant to understand tasks, generate responses, and execute workflows.

It’s worth noting that you can also connect OpenClaw to a local AI model, allowing complete privacy, customization, and cost-free setup.

How OpenClaw works

OpenClaw is built around automation instead of conversation. After installing the platform and linking it to your preferred AI model, you expand its capabilities through skills, which act as integrations with external services. You then define jobs, which are persistent instructions describing what the assistant should do and when it should do it.

Once configured, OpenClaw runs continuously in the background, executing tasks and delivering results without requiring you to return and reissue prompts.

How your data is handled

The data processing is split between your local hardware and the cloud. Your computer acts as the “Body” that keeps your files, conversation history, and API keys stored privately on your hard drive, while the external AI (like OpenAI or Anthropic) acts as the “Brain” that receives only the specific text snippets needed to reason through your request and send back instructions for your computer to execute.

Real-world example

A practical use case highlights its value. You can connect your Gmail account using a Google Workspace integration and create a job that sends you a WhatsApp notification whenever a specific person emails you. The notification can include a summary of the message and even a suggested reply.

Instead of constantly checking your inbox or prompting a chatbot for updates, the assistant monitors activity and handles the workflow automatically.

Why OpenClaw is different

The key difference between OpenClaw and tools like ChatGPT or Claude is persistence. Traditional AI tools require active sessions in which you continuously provide input, whereas OpenClaw runs independently once configured.

It integrates with services such as Gmail, Slack, GitHub, Telegram, WhatsApp, and Discord, allowing it to operate across your existing ecosystem. Because it is self-hosted, you maintain full control over permissions, data access, and behavior.

In this guide, I’ll outline the steps to install OpenClaw on Windows 11.

Install OpenClaw on Windows 11 via PowerShell
Install OpenClaw on Windows 11 via WSL

Install OpenClaw on Windows 11 via PowerShell

To install OpenClaw on Windows 11, follow these steps:

Open Start on Windows 11.
Search for PowerShell (or Terminal), right-click the top result, and select the Run as administrator option.
(Option 1) Type this command to install OpenClaw on Windows 11 and press Enter:
```
iwr -useb https://openclaw.ai/install.ps1 | iex
```
Select “Yes” to confirm the security access and press Enter.
Choose the QuickStart option for the Setup mode and press Enter.
Select the AI model provider, such as OpenAI or Anthropic, and press Enter.
Choose the authentication method to connect to the API key, then press Enter.
Confirm the API key for the AI model provider and press Enter.
Choose the default AI model from the provider and press Enter.
Use the default gateway port and press Enter.
Use the default gateway bind and press Enter.
Use the default gateway auth and press Enter.
Use the default for tailscale exposure and press Enter.
Press Enter to generate a token.
Select your channel, such as Telegram, WhatsApp, and others.
Configure the skills (as necessary).
(Optional) Install missing skills dependencies (as necessary).
Skip or configure additional AI model providers.
Enable all the different OpenClaw hooks (as necessary).
Choose “Yes” to install the OpenClaw gateway service.
Select “Node” to enable the gateway service runtime.
Press Enter to enable bash shell completion for OpenClaw.
Continue with the on-screen directions to complete the setup.

Once you complete the steps, the command will take care of the entire installation. It’ll perform a system check, install Node.js via the system package manager (if it’s missing), download the latest OpenClaw build, and immediately start the interactive Onboarding Wizard to link your OpenAI or Anthropic keys.

Install OpenClaw on Windows 11 via WSL

To install OpenClaw using the Windows Subsystem for Linux, install WSL on Windows 11, and follow these instructions:

Open Start.
Search for and open your WSL setup, such as Ubuntu.
Type this command to install OpenClaw on Windows 11 and press Enter:
```
curl -fsSL https://openclaw.ai/install.sh | bash
```

Select “Yes” to confirm the security access and press Enter.
Choose the QuickStart option for the Setup mode and press Enter.
Select the AI model provider, such as OpenAI or Anthropic, and press Enter.
Choose the authentication method to connect to the API key, then press Enter.
Confirm the API key for the AI model provider and press Enter.
Continue with the on-screen directions to complete the setup.
Choose the default AI model from the provider and press Enter.
Use the default gateway port and press Enter.
Use the default gateway bind and press Enter.
Use the default gateway auth and press Enter.
Use the default for tailscale exposure and press Enter.
Press Enter to generate a token.
Select your channel, such as Telegram, WhatsApp, and others.
Configure the skills (as necessary).
(Optional) Install missing skills dependencies (as necessary).
Skip or configure additional AI model providers.
Enable all the different OpenClaw hooks (as necessary).
Choose “Yes” to install the OpenClaw gateway service.
Select “Node” to enable the gateway service runtime.
Press Enter to enable bash shell completion for OpenClaw.
Continue with the on-screen directions to complete the setup.

After you complete the steps, the command will check and install the necessary components to run OpenClaw in the Linux container.

OpenClaw system requirements for Windows 11

The requirements depend on whether you are using external cloud APIs or running the AI models locally on your own hardware.

Software requirements

OS: Windows 11 22H2 or higher recommended.
WSL: Latest version highly recommended for stability, as the core agent logic is Linux-optimized.
Runtime: Node.js 22.14+.
Execution Policy: PowerShell must be set to RemoteSigned to run the installation scripts.

Hardware Requirements

Because OpenClaw is a persistent agent, it’s more memory-heavy than a standard chat app.

CPU: Modern Dual-core (Intel i5/Ryzen 5).
RAM: 8GB (minimum).
Storage: 4GB of free storage.

These are the hardware recommendations for local AI models:

CPU: NPU-enabled CPU (such as Ryzen AI or Intel Core Ultra).
RAM: 16GB to 32GB (Crucial if running Ollama or Llama 4).
Storage: 40GB+ of free storage.

Security and OpenClaw

On the other hand, if you used the second option, the software is installed but unconfigured. If you try to run openclaw immediately, it’s likely to throw an error because it doesn’t have an API key (OpenAI or Anthropic) to “think” with.

OpenClaw fundamentally changes how you use AI on Windows 11 by shifting from reactive prompting to proactive automation. Instead of constantly interacting with a chatbot, you deploy a persistent assistant that runs in the background and executes tasks on your behalf.

However, similar to other AI agents and features such as Windows Recall, this approach raises legitimate security concerns.

OpenClaw’s autonomy is what makes it powerful, but it also raises the stakes. Once connected to services such as your email, calendar, or file storage, the assistant has real access to your data. This is not a sandboxed chatbot. It is an active system with permissions to read, trigger, and act across your accounts.

That level of access means your security posture matters. If API keys or access tokens are exposed, an attacker could take control of your setup, execute malicious workflows, and potentially cause data loss, service abuse, or financial damage.

The safest approach is to start small and expand gradually. Only connect the services you truly need, carefully review permissions, and take the time to understand how each integration behaves before relying on it for critical tasks.

Most importantly, treat your API keys and tokens like passwords. Store them securely, never expose them publicly, and rotate them regularly to reduce risk.

Thank you for your feedback!

Tell me why, so I can fix it