Environments

Where the agent acts, such as a cloud browser, your Chrome, or your desktop.

Where the agent acts

Your browser

Chrome on your machine, with your logins. web on user_device.

Your desktop

Your screen, mouse, and keyboard. desktop on user_device.

Cloud browser

A browser H hosts per session. The default. web on cloud.

How it works

The agent works one step at a time: look at the screen, reason, act. Each environment gives the agent a fixed set of actions plus guidance on using them. The browser, for example, adds navigate, click, type, scroll, and read. To shape how the agent uses them, set the agent’s instructions or attach skills. To add your own functions, use custom tools.

Pre-built environments

H ships browsers in the h/ catalog. List them with GET /api/v2/environments.

Create your own

Most agents define their environment inline, in the agent itself:

Python

from hai_agents import Client

client = Client()
result = client.run_session(
    agent={
        "name": "wide-web",
        "description": "Browses with a full HD viewport.",
        "environments": [{"id": "wide-browser", "kind": "web", "mode": {"type": "visual", "width": 1920, "height": 1080}}],
    },
    messages="Open hcompany.ai and describe the page header.",
)
print(result.answer)

When several agents share one, create it in your catalog and reference it by id: "environments": ["wide-browser"]. The Cloud browser page lists every browser field.