ego-browser

The browser tools your AI agent uses with ego (lite).

ego-browser lets an AI agent work in ego (lite). It can open a page, read a Snapshot, use the page's controls, and check the result in a Space.

How an agent uses it

For a typical task, the agent:

Opens or returns to a Space for the task.
Opens the page it needs.
Reads a Snapshot to find the relevant text and controls.
Takes an action, then reads the page again to check what changed.
Returns the result with links or other details you can verify.

The tools are JavaScript helpers. This short example shows how an agent opens a page and reads it:

await useOrCreateTaskSpace('check a page')
await openOrReuseTab('https://example.com', { wait: true })
cliLog(await snapshotText())

Your agent runs the code through the ego-browser skill. You can simply describe the task in the agent's chat. Quick start has a prompt you can try.

Decide when to step in

Ask the agent to pause for sign-ins, verification codes, payments, or changes you want to approve. You can open its Space to handle that step and review the result afterward.