Agent SDKOntology SDKConsole

Let your agent use its own browser

Give a permitted workspace agent a persistent browser tab per chat, with its deployed settings visible in the console.

The workspace agent can now use its own cloud browser to work with sites that need JavaScript or sign-in. Its profile belongs to the agent, not the person chatting with it, and each chat keeps its own tab. The console's Agent page has a Browser tab showing the deployed setting, proxy and recent sessions.

For example, ask the agent to check the status of a request in a signed-in vendor portal and report what it finds. It can keep its own sign-in between chats without asking a teammate to share their browser session. Or ask it to test a JavaScript-heavy signup flow: open the page, click through the steps and inspect screenshots to spot a broken state before you ship.

Turn it on in agent/segmentstream.json alongside the sandbox:

{
  "sandbox": true,
  "browser": {}
}

Grant the caller's role agent_tool:browser and agent_tool:sandbox. The agent runs short Python browser scripts and sees only what they print; screenshots can be read separately. Direct traffic is the default unless the workspace config chooses a residential proxy. Guests cannot use the tool. Browsers are replaced after their session limit, while the agent's saved profile carries sign-ins forward. The console does not yet show recordings.

Build on SegmentStream.

Start a pilot

Our engineers build the first version on your data in two weeks, at a fixed price.

Or build it yourself with your coding agent:

curl -fsSL https://segmentstream.com/install.sh | sh

We’ll email you product updates, tips and onboarding help. Unsubscribe anytime.