Create Your First Hosted Agent

Before you create an agent, decide what work it should own. The clearer the job is, the easier it is to configure the instructions, workflows, and access it needs.

Examples include research support, follow-up drafting, customer support triage, recruiting coordination, or recurring operational tasks.

When you create an agent, you choose the hosted runtime plan and any workspace storage you need. That defines the cloud environment the agent will run inside.

The point is to start with enough capacity for the work without overcomplicating the setup.

Once the runtime exists, configure how the agent should behave, what priorities it should follow, and what files or tools it should use.

This is where Open Vassal becomes more than hosting. It gives you the operational surface to shape an agent into something specific and repeatable.

A good first setup usually includes three things: what kind of work the agent owns, how it should communicate, and what information or files it should rely on by default.

If the setup still feels vague after that, the problem is usually not that you need more features. It is usually that the task boundary still is not sharp enough.

After the initial setup, run a few realistic tasks all the way through. Look at the output quality, the missing context, and where the workflow still depends on manual cleanup.

This is the fastest way to figure out whether the runtime plan is large enough, whether the workspace needs more files, and whether the instructions are actually usable.

The first version does not need to be perfect. It just needs to handle one real slice of work reliably enough that you know what to improve next.