How to Run OpenClaw on an Acurast Processor
OpenClaw is an open-source personal AI assistant, the kind that doesn’t just answer questions but actually goes and does things: running commands on a schedule, generating summaries, analysing your email, and carrying out multi-step tasks on your behalf.
The catch is that an assistant like that is only useful if it’s actually running. Close your laptop and a locally installed OpenClaw stops working. That’s why most people end up renting a VPS or a cloud instance just to keep it online, and why “best VPS for OpenClaw” has become such a common search.
There’s another option. Through Acurast Cargo, our collection of ready-to-deploy applications, you can self-host your own OpenClaw instance on the Acurast network without building the deployment setup from scratch and without renting a server from anyone.
Your instance runs on an Acurast processor, a smartphone that someone has dedicated to the network and that earns rewards for running workloads like yours. It stays available whether or not your own computer is switched on, which is exactly what you want from an assistant that needs to execute scheduled tasks, process information, or stay reachable while you’re away.
In this tutorial, we’ll configure the OpenClaw Cargo app, deploy it on Acurast, and sign in to the OpenClaw dashboard.
Why Self-Host OpenClaw on Acurast Instead of a VPS
The usual answer to “where do I run OpenClaw?” is a rented box: a VPS, a cloud instance, a spare Mac mini, or a Raspberry Pi humming away in a cupboard. Each of those works. Each also comes with a monthly bill and a machine you have to keep patched.
Acurast Cargo takes a different route. Cargo runs full Linux-based workloads on decentralized compute, which means existing software doesn’t have to be rewritten to run there. OpenClaw installs from npm on an Acurast processor in much the same way it would install on a server you rented.
This is the same idea behind running a WordPress site, S3-compatible object storage or a Minecraft server on Acurast. The workload doesn’t need to know the infrastructure underneath it has changed.
For an AI assistant specifically, there’s a second reason this matters. An agent that only thinks while you’re watching it isn’t much of an agent. We wrote about that problem in why your AI shouldn’t stop working when you close your browser. Putting OpenClaw on a processor gives it somewhere to keep running.
Prerequisites
Before getting started, have these ready:
- The Acurast example apps repository (linked in step 1)
- An Acurast mnemonic — the seed phrase for the account that pays for the deployment
- An SSH password — SSH is the standard way to get a command line on a remote machine
- A callback URL — a webhook endpoint that receives the deployment’s status messages
- An OpenRouter API key — OpenRouter is a service that gives you access to models from many providers behind a single key
- A password for the OpenClaw web interface
Once you have those, you can start setting up the deployment.
Self-Hosting OpenClaw Step by Step
Step 1: Open the OpenClaw Cargo App
Start by checking out the Acurast Example Apps repository.
Navigate to the OpenClaw application folder, app-cargo-openclaw.
This is the OpenClaw app included as part of Cargo. The directory contains everything needed to deploy OpenClaw on Acurast, including the setup script for OpenClaw itself and the tunnel that makes the deployment reachable from outside the processor.
Before deploying, you’ll need to configure a few environment variables.
Step 2: Configure the Environment
Copy the provided environment file (.env.example to .env) and fill in the required values.
Acurast mnemonic. First, set your Acurast mnemonic. This is the seed phrase for the account used to pay for the deployment on the Acurast network. Because a mnemonic provides access to funds, keep it private and never commit your populated .env file to a public repository.
SSH password. Next, configure an SSH password. SSH access lets you connect directly to the processor if you need to debug the deployment, update OpenClaw, or run the openclaw command line tool. Use a strong, unique password, particularly because the deployment is remotely accessible.
Callback URL. Once the deployment finishes, a unique URL is generated for your OpenClaw instance, and we need somewhere to receive it. For this example we use a webhook service. Any service that can receive POST requests and show you their contents will do. Open the service, create or copy your personal endpoint, and add it as the callback URL in your configuration. When the deployment sends its request, the generated OpenClaw URL will appear there.
Network. Leave the Acurast network set to mainnet, where most processors are available.
Step 3: Configure a Custom Domain (Optional)
The configuration also provides an option for using your own domain.
Depending on the current deployment setup, this may no longer be necessary, because the app falls back to a default Acurast domain when no custom suffix is set. If you don’t need a custom domain, simply leave the default configuration in place.
If you do want to use one, follow the relevant Acurast domain configuration instructions before deploying.
Step 4: Add Your OpenRouter API Key
OpenClaw needs access to an LLM (a large language model), so the next step is to configure your OpenRouter API Key.
Add your API key to the environment configuration and select the default model you want OpenClaw to use. The example configuration may specify an older model. That’s adequate for testing the proof of concept, but for a longer-running deployment you may want a newer model chosen on performance and cost.
As with your mnemonic and passwords, treat the API key as a secret and avoid committing it to source control.
Step 5: Set the OpenClaw UI Password
You’ll also need to set a password for the OpenClaw web interface, which is what protects your assistant’s public URL from anyone who stumbles across it.
You’ll use this password later when opening your deployed instance, so keep it somewhere secure.
At this point, the main environment configuration is complete.
Step 6: Review the Acurast JSON Configuration
Before starting the deployment, take a look at the acurast.json configuration file.
One of the settings you can change here is the deployment duration, which determines how long your OpenClaw instance will run. Choose it carefully, because the deployment is temporary: once its configured runtime ends, the deployment stops and its data is deleted.
You should also leave the minimum Android version at its provided value. This helps ensure your deployment is assigned to a processor that supports all the features OpenClaw requires.
Step 7: Deploy OpenClaw
With everything configured, start the deployment.
The Acurast CLI registers the deployment and displays the associated execution price. During deployment, you may see a message indicating that the selected fee is higher than necessary. In that case, you can choose the suggested fee instead.
The deployment then waits for a suitable processor to pick it up. This should take a few minutes.
Step 8: Check the Deployment Logs
Once the deployment begins, follow the logs at your callback endpoint to see the environment being prepared.
Among other things, the deployment installs and configures:
- Dropbear, a lightweight SSH server
- The reverse tunnel that makes the instance reachable from the public internet
- Node.js
- OpenClaw itself
Eventually, the logs should indicate that OpenClaw is ready. You should also receive the unique URL assigned to the deployment through the callback endpoint you configured earlier.
Step 9: Open the OpenClaw Dashboard
Copy the generated deployment URL and open it in your browser.
You’ll be prompted for the UI password you configured in the .env file. Enter it, and the OpenClaw dashboard should load.
That’s it: you now have your own private, self-hosted OpenClaw instance running on an Acurast processor through Cargo, and nothing rented to put it on.
Step 10: Start Using OpenClaw
From the dashboard, you can start a new session and interact with your OpenClaw instance. Depending on your configuration, OpenClaw can schedule actions, generate summaries, work through information, and carry out other AI-assisted workflows.
If you encounter tool-call errors during initial testing, check whether those particular tools are actually required for the task you’re trying to perform. Basic model interactions can still work even when an optional tool isn’t configured correctly.
Once you receive a response from the model, you’ve confirmed that the core OpenClaw deployment is working.
Important: Your Deployment Is Temporary
There’s one important detail to remember when self-hosting OpenClaw this way.
The deployment only runs for the duration you specified in the Acurast configuration. Once that deployment ends, its data is deleted — including OpenClaw’s configuration and its memory of your sessions.
Choose a deployment duration that gives you enough time to use or test your instance, and don’t treat the processor’s local storage as permanent storage for anything you need to preserve.
Watch the Full Video Tutorial
Want to see the entire setup in action? The video below walks through the whole OpenClaw configuration and deployment process step by step, from filling in the environment file to signing in to the dashboard.
Wrapping Up
By making OpenClaw available through Cargo, Acurast gives you a straightforward way to self-host your own AI assistant on the network.
The overall process is simple: configure the OpenClaw Cargo app, choose your deployment settings, deploy to Acurast, retrieve your instance URL, and sign in to OpenClaw.
Once everything is running, you have your own OpenClaw instance hosted on an Acurast processor, ready to handle AI-powered tasks for the lifetime of the deployment — without a cloud server anywhere in the picture.
Curious about the phones doing the work? Any spare Android device can become an Acurast processor and earn rewards for running deployments like this one. And if you want to build your own Cargo app, start with the developer resources.
