Launch OpenClaw.Keep your AI online.

OpenClaw is an open-source AI assistant that connects a model to your chosen channels and tools. Its Gateway coordinates messages, tasks and sessions in a workspace you configure.

You decide which model it uses, what it can access and which actions should require review.

The listed plans include a preconfigured OpenClaw workspace, automated installation, one-click updates and dedicated support. Starter, Pro and Business differ in CPU, RAM, SSD storage and bandwidth.

You still connect your model provider, messaging channels and the accounts needed for your tasks. Compare the current plan allocations before ordering.

The hosting price covers the listed OpenClaw environment; model usage is separate. Connect a model provider or compatible endpoint and check its pricing and limits before you automate frequent jobs.

Start with one task, observe its usage, then add schedules or more agents when the cost is predictable.

Choose by the work the server will run, not just the number of prompts you expect to send:

• Starter: 1 vCPU and 2 GB RAM for a lighter assistant using a remote AI model.
• Pro: 2 vCPU and 4 GB RAM for more tools or overlapping work.
• Business: 4 vCPU and 8 GB RAM for heavier browser tasks or several active workflows.

Browser automation and concurrent jobs can use more memory than simple chat. Watch resource usage as you build.

Ask it to research and draft a brief, organise information from connected apps, prepare code for review or run a scheduled check. The useful task is the one for which you have supplied the right tool, access and instructions.

For example, connect a calendar and a messaging channel, then ask for a morning meeting summary. Review the result before making that task recurring. See more examples.

Yes. The software is open source under the MIT licence; the licence itself has no subscription fee. Hosting, model calls and any paid integrations are separate costs.

Truehost’s plan price is for the hosted environment, not an unlimited model allowance. See the OpenClaw licence.

After the prepared installation is available, work through one small task:

1. Connect a supported AI model provider or endpoint.
2. Choose a channel or use the OpenClaw interface.
3. Give the assistant only the tools and accounts that task needs.
4. Send a test request, check the output and adjust permissions before scheduling it.

OpenClaw’s getting-started guide covers the current setup flow.

The Gateway does not need a GPU when your AI model runs with a cloud provider; the provider handles inference. These Truehost plans are not advertised as GPU servers.

If you want to run a model on the same machine, size the hardware for that particular model first. Its memory and accelerator needs are separate from the Gateway’s needs.

OpenClaw’s basic hosting guidance lists 1 GB RAM as an absolute minimum and 2 GB or more as a more practical starting point for the Gateway. Browser work, media processing and concurrent tasks need more headroom.

Leave storage for the operating system, dependencies, logs and workspace files. The guidance’s small installation figure is not a complete disk budget, especially if you add a browser or local model.

Yes. Local use can suit learning and tasks that need files or devices on your computer. That machine must stay awake and connected for scheduled or ongoing work.

A hosted Gateway keeps server-based tasks running independently of your laptop. A task that relies on a paired local device still needs that device online. Compare the local and hosted options.

For a dedicated Linux server, a current Debian or Ubuntu release is a straightforward path in OpenClaw’s installation guidance. For a local setup, choose the supported system you can maintain and that runs your required tools.

Check the current installer and dependency requirements before moving an existing workspace between operating systems.

Server-based tasks can continue while your laptop is off because the Gateway runs in the hosted environment. The server, model provider and connected services still have to be available.

The VPS cannot reach private laptop files or applications unless you deliberately connect a supported local device, and that device remains online.

OpenClaw supports channels such as Telegram and other integrations you configure. Connect the account or bot, set who may send it messages and test a simple request before giving it access to more tools.

A channel is an entry point, not a blanket permission to every account in your workspace. Check the current channel options for the setup you want.

Yes, if you enable and configure a browser tool. OpenClaw can use an agent-controlled browser profile to read pages and interact with them. Test the workflow on a low-risk task before allowing account changes or submissions.

Keep browser access and credentials scoped to the work. The browser guide explains the available modes.

A VPS gives the Gateway a persistent home; permissions and exposure determine what it can do. Keep access private or authenticated, limit connected tools, and review third-party skills before installing them.

• Give each workflow only the accounts and capabilities it needs.
• Require a person to review consequential actions such as publishing or payments.
• Keep updates and recoverable backups in your operating routine.

Use OpenClaw’s security guide when configuring access.

Yes. Back up the workspace, configuration and state, restore them on the new host, and test the result before retiring the old installation.

Check credentials, file paths, channels and scheduled jobs after the move. Follow the current migration steps.

The catalogue lists one-click updates and dedicated support. Before relying on recurring work, establish who applies application and system updates and who maintains recoverable copies of your workspace and settings.

Back up before changes, test a restore, then check model access, a channel message and a scheduled task after updating.

You can build multiple workflows in one workspace, but simultaneous browser sessions and tool calls compete for CPU and memory. Start with one dependable workflow and measure its resource use before adding parallel work.

If jobs queue or browser tasks slow down, compare the larger Pro and Business allocations.