O1 / Browser assistant

Work in a tab. Review the next move.

Select a Chrome tab and review proposed steps while staying in control.

Explore the public preview. Live AI requires a verified account; supported workflows depend on enabled tools and account limits.

See how it works

Browser assistantWORKFLOW WALKTHROUGHExample

EXAMPLE REQUEST

In this selected tab, open the visible contact section and prepare a draft inquiry asking about workshop availability. Use only the text I provide. Let me review each proposed action. Stop before submitting the form or sending a message.

Illustrative action review

Context: The tab you selected, with its site and page title visible in the task.
Proposed action: Click the visible Contact link, then review the newly loaded page before another step.

More example details

Draft: Fill a reviewed inquiry field with the text you supplied. Keep submission for your manual review.
Receipt: Confirm whether the action applied. A proposal, connection error, or stale task must not be treated as completed work.

01 / THE WORKFLOW

What can o1 help you do?

The o1 browser assistant works through the Chrome extension and a tab you select. Give it a bounded task, inspect the visible control it proposes to use, and review the next step. The connection and selected-page permissions are part of the workflow; a URL in ordinary chat does not grant control of your browser.

Start with a low-risk task you can inspect, such as opening a visible section, choosing an option, or filling a draft field with text you supplied. The assistant proposes steps using the controls it can identify on the selected page. A request to click is distinct from confirmation that the page applied the click.

02 / THE BRIEF

Start with a specific prompt

Use this as a starting point. Replace the example details with your own task and constraints.

YOUR REQUEST
In this selected tab, open the visible contact section and prepare a draft inquiry asking about workshop availability. Use only the text I provide. Let me review each proposed action. Stop before submitting the form or sending a message.

This prompt is an example. Its outcome depends on the tools enabled in your workspace and the information you provide.

03 / GET STARTED

How to use the browser assistant

01
Connect Chrome
Sign in to o1 and open the Chrome extension connection flow. Follow the extension setup and grant the selected-page permissions you intend to use.
02
Choose the tab
Select the relevant tab, check its title and site, and describe one task. Keep the page context visible so an unrelated tab is not mistaken for the target.
03
Review the exact action
Inspect the control, text, or destination in the proposed step. Approve only if it matches your request. If the control is missing or unclear, refresh the page context instead of guessing.
04
Check the outcome
Watch the task receipt and selected tab. If the page changes or the connection becomes stale, review the current tab and retry before continuing.

Already have an account? Open the Chrome connection flow.

04 / YOUR CONTROL

Permission and continuity belong in the task

The selected tab is the scope

Always verify the current title and site. Moving to another tab or a changed page requires fresh context before the next action.

Sensitive steps need your judgment

Keep purchases, credentials, payments, messages, and publishing under direct review. The assistant may decline risky or ambiguous controls.

Check the actual receipt

The page may load slowly, reject an input, or move a control. Confirm the applied result before treating the task as done.

05 / CURRENT SCOPE

Know the limits before starting

The public demo shows a workspace preview. It cannot browse, open tabs, click controls, or send messages; live AI requires a verified account.
Browser actions require the Chrome extension, a selected tab, and appropriate permissions. Some pages and controls are unavailable or unsupported.
Web search results and shared page text are evidence for an answer, not proof that a browser action ran. o1 does not claim success without an action outcome.

Check the workspace's current availability and account limits before starting a job. Your chat and connected tools may have different capabilities from the public demo.

06 / QUESTIONS

Before your first task

Can ordinary o1 chat control Chrome?

Browser control uses its separate Chrome action flow. A chat answer, web search, or pasted URL does not mean a tab was opened or changed.

What happens when the selected tab changes?

Review the current tab and refresh the task context before continuing. Confirm the new site and page title so the next proposed step has the right target.

Can I try this in the demo?

You can explore the public workspace preview and illustrative action walkthrough. Live AI requires a verified account. Actual page actions also require the extension connection and selected-tab permissions in the signed-in workflow.

GUIDES & TEMPLATES