There are two ways to automate HyperView: give an AI agent the hv command-line tool, or call the REST API from your own code. Both use the same API key, the same kind of session, and the same capability rules.
The two paths
An AI agent with the hv CLI. Install the CLI where your agent runs, sign it in, and give the agent the HyperView skill. Then ask in plain English, such as take a screenshot of windesktop12 or install LibreOffice on Linux1234, and the agent works out the commands. Set it up from the AI Agents screen.
The REST API. Your code exchanges an API key for a short-lived token, finds a system, opens a session, acts on it and closes it. It's standard, versioned REST with no SDK needed. Set it up from the Integration screen.
Which to use
Use an AI agent when you don't know the steps in advance:
Exploring or debugging a machine you can't easily reach
A one-off task across your fleet
Use the REST API when the steps must be identical every time:
A repeatable step in CI
End-to-end tests of a desktop application
A feature inside your own product
A fleet task you run nightly
An agent is good at working out how to do something. Code is good at doing it the same way a thousand times.
What they have in common
Both need an API key with the
USE_AUTOMATIONpermission.Both open an automation session, billed exactly like a person's session.
Both are limited by the same capability policies and the same account-wide Automation Enabled switch.
The hv CLI is built on the REST API, so anything the CLI does, your code can do too. You can mix them: script the predictable parts of a workflow and let an agent handle the rest.
Next steps
Automation capabilities: read this before you give either one access