Skip to content
Docs
Docs/Build

Work with the browser agent

Follow a browser task in RNBlocks, take control for private steps and return to the same conversation.

4 min readUpdated Oct 8, 2026

The browser agent is in beta. Some websites, sign-in steps and handoffs may not work smoothly yet.

The agent can use a browser to help with a task. You can follow its work and take control when a step needs you, such as a private sign-in or account choice.

On this page

Give the browser a specific task

The browser card shows the current session. Expand it for a larger view. A finished session is the last view of that task, not a browser the agent is still controlling.

Complete a private step

  1. Read the handoff request

    Check the site and what the agent needs you to do. Keep account passwords, verification codes and payment details out of the conversation.

  2. Pause the agent or open its handoff

    If the agent is still browsing, select Take control and wait for it to pause. If it has already asked you to complete a private step, go directly to Take over browser.

  3. Complete the step yourself

    Select Take over browser to open the interactive browser. Use Expand or the zoom controls if needed, then complete the requested action.

  4. Return control to the agent

    Select Return browser control to agent inside the browser. The conversation resumes so the agent can check the outcome; you do not need to resend the original request.

Take over the browser when needed

Example: a browser handoff in Pocket Chef’s Studio conversation. Illustrated walkthrough.

  1. Read the browser task and the handoff request.
  2. Choose Take over browser to complete the step yourself.
  3. Return control when you are ready for the agent to continue.
Take over the browser when needed

Example: a browser handoff in Pocket Chef’s Studio conversation. Illustrated walkthrough.

When a session closes or disconnects

If the control reports a disconnection, use Take over browser again when offered. If the session has ended, tell the agent what completed and ask it to check the current provider state before repeating an action.

A sign-in, payment or provider verification is complete only when the relevant service confirms it. A browser screenshot or a guide is not confirmation by itself.

Documentation
Open Studio