Skip to main content
1

Get your API Key

Visit the Dashboard’s Settings tab:
Copy your API key for the next step.
2

Install Agent Browser

The install command downloads Chromium for local use. When using the Browserbase provider, the browser runs in the cloud instead.
3

Set environment variables

Set your Browserbase credentials:
4

Open a page with Browserbase

Use the -p browserbase flag to run commands on a Browserbase cloud browser:
This creates a Browserbase session and navigates to the URL. You can view the session in your Browserbase Dashboard.
5

Get an accessibility snapshot

The snapshot command returns an accessibility tree with element references. This is the primary way AI agents understand page content:
Output includes element refs that you can use in subsequent commands:
Use the -i flag to show only interactive elements:
6

Interact with elements

Use element refs from the snapshot to click, type, and fill:
7

Close the session

All commands support the -p browserbase flag to run on Browserbase cloud browsers. You can also set AGENT_BROWSER_PROVIDER=browserbase as an environment variable to avoid passing the flag each time. See the Agent Browser README for the full command reference.

​
Connecting via CDP

You can also connect Agent Browser to a Browserbase session directly using the CDP WebSocket URL:
Node.js
Then connect Agent Browser:

​
Expected output

The open command navigates the cloud browser, and the snapshot command prints an accessibility tree with element references. The available references depend on the current page; use the latest snapshot when selecting an element.

​
Cleanup

Use the close command from the final setup step to end the browser connection. If you created a session separately through the SDK, also release it using Manage browser sessions. Unset the Browserbase API key in the terminal when you finish testing.

Agent Browser GitHub

Full documentation, examples, and source code.

Browserbase sessions API

Learn more about creating and managing Browserbase sessions.