How to Choose the Right Instance Size for Your OpenClaw Agent on Cloudways

Choosing the right instance size is important because it helps your OpenClaw Managed AI Agent run smoothly based on the work it needs to handle.

The instance size you select controls the amount of server resources available to your OpenClaw installation, such as CPU, memory, and storage.

When selecting an instance size, the most important factor is not the size of your business or the number of people in your team.

The most important factor is how many sub-agents may run at the same time inside your OpenClaw installation.

A sub-agent is a smaller agent or workflow inside your main OpenClaw setup that performs a specific task, such as handling support requests, generating content, running summaries, or managing internal operations.

If your agent only runs one simple workflow occasionally, a smaller instance may be enough. If your agent runs multiple workflows throughout the day, connects to customer-facing channels, or has several sub-agents working at the same time, you should choose a larger instance.

Cloudways provides different OpenClaw instance tiers based on CPU, memory, and storage. CPU helps process tasks, memory helps the agent handle active work, and storage keeps the data and files related to your OpenClaw installation.

Each tier is designed for a different level of workload and concurrency. Concurrency means how many sub-agents or workflows are active at the same time.

Scout includes 1 vCPU, 2 GB memory, and 50 GB storage.

Scout is suitable for simple and low-activity OpenClaw setups. It is a good option for personal assistants, basic daily summary jobs, testing, learning, and early prototypes.

A prototype is an early version of your setup that you use to test an idea before using it in production.

Scout works best when you have one or two lightweight sub-agents and they do not run at the same time.

For example, you can use Scout if your agent only creates a daily report, summarizes basic information, or performs a small scheduled task.

You should avoid Scout if your OpenClaw agent handles live customer-facing channels, such as support or chat channels, where activity can increase suddenly.

These types of workloads can be unpredictable and may need more resources than Scout provides.

Operator includes 2 vCPU, 4 GB memory, and 80 GB storage.

Operator is a good starting point for most first-time customers who want to run a single production OpenClaw agent. A production agent is an agent that is being used for real tasks, not just testing.

This tier is suitable when your OpenClaw setup has one to three sub-agents, moderate workflow activity, and one connected channel under normal load. It can also support a setup where three to five sub-agents are defined, but only one or two are active at the same time.

Operator is a practical choice when you are launching your first OpenClaw agent and expect regular usage, but not heavy parallel activity.

For example, it can work well for a single internal assistant, a controlled support workflow, or an agent connected to one main channel.

Squad includes 4 vCPU, 8 GB memory, and 160 GB storage.

Squad is designed for OpenClaw setups where multiple sub-agents need to work together.

This tier is suitable for customer service workflows, content generation workflows, and internal operations where more than one task may run at the same time.

You should consider Squad if you have multiple channels connected, workflows running throughout the day, or several sub-agents working in parallel.

It is suitable for setups with six to ten sub-agents, where three to five sub-agents may be active at the same time.

For example, Squad may be a good choice if one sub-agent handles customer questions, another creates summaries, another updates internal records, and another checks information from connected tools during the same period.

Swarm includes 8 vCPU, 16 GB memory, and 320 GB storage.

Swarm is designed for high-concurrency workloads. This means it is suitable when many sub-agents or workflows may run at the same time.

It is also suitable for larger OpenClaw installations, large sub-agent fleets, and agencies that manage automation for multiple clients from one OpenClaw instance.

You should consider Swarm if your setup has more than ten sub-agents and more than five may be active at the same time. This tier provides more resources for heavier workloads and busy environments.

However, if your workload is very large or serves different clients, you should also consider whether using more than one OpenClaw instance would provide better separation.

For example, an agency may prefer separate instances for separate clients to keep workloads isolated and easier to manage.

Use the following guidance to choose your instance size:

As a simple rule, Scout is best for learning and testing. Operator is the safest default for most first-time production users.

Squad is better when multiple workflows run during the day. Swarm is best for larger and busier setups.

You may need a larger instance size if your OpenClaw setup starts showing performance issues during normal or peak activity.

One sign is that the OpenClaw container restarts unexpectedly during busy periods.

This may happen when the instance does not have enough memory to handle the active workload.

Another sign is that workflows take longer to complete even though you have not changed the workflow itself.

This can indicate that the instance is under pressure because too many tasks are running or the workload has grown.

You may also notice that the OpenClaw dashboard feels slow while sub-agents are running.

If the dashboard becomes sluggish during active workflows, the current tier may not have enough resources for your usage pattern.

If these signs appear often, you may have outgrown your selected tier.

The instance tier affects the resources available to your OpenClaw installation, but it does not affect every part of your setup.

It does not affect your LLM costs. LLM costs are billed separately by your AI model provider.

An LLM, or Large Language Model, is the AI model used by your agent to understand and generate responses.

The instance tier also does not affect the region or latency choice. Region selection is handled separately when you launch the instance. Latency means the time it takes for data to travel between the user, the server, and connected services.

The instance tier also does not change feature availability. All tiers include access to supported OpenClaw features such as SSH, backups, channel integrations, and BYOK.

BYOK means Bring Your Own Key, which allows you to use your own API key for supported AI providers.

Because resizing is not available during Public Preview, you should choose your starting tier carefully.

Think about how many sub-agents you plan to create and how many of them may run at the same time during your busiest period.

If your agent will only be used for testing or simple scheduled tasks, Scout may be enough.

If your agent will support real users or connect to live channels, Operator is usually a better starting point.

If your workflows are expected to run frequently throughout the day, Squad may be more suitable.

If you expect high activity, many sub-agents, or agency-level usage, Swarm may be the better option.

When you are unsure between two tiers, choose the higher tier. This helps reduce the chance of performance issues after launch and avoids the need to create a new instance sooner than expected.

The right OpenClaw instance size depends mainly on concurrency, which means how many sub-agents run at the same time.

Scout is suitable for simple testing and learning. Operator is the recommended default for most first-time production setups.

Squad is better for multiple workflows running during the day. Swarm is suitable for high-concurrency workloads and larger installations.

Choose your tier based on your expected active workload, not only the number of sub-agents you plan to create. Since resizing is not currently available during Public Preview, plan ahead and select a tier that can support your expected usage.

That’s it! We hope this article was helpful.

If you need assistance, feel free to:

We're here 24/7 to help you!