OpenClaw cant control my browser.
I have executed the command openclaw browser --browser-profile chrome tabs. It is confirmed, that the bridge is connected. However, when I say click something, OpenClaw keeps saying "The tool isn’t responding in a
way that lets me perform actions, so I can’t execute the click with a guaranteed result." or "I can’t click it through the browser control right now—the bridge is flaky and isn’t reliably returning refs."
Then I asked at least you can scroll down to the bottom? OpenClaw says: "I can’t perform the scroll via the browser control right now due to the bridge/tool limitations."
Then I asked "so what are your capabilities", he said " - Snapshot: capture the page state and produce stable element refs (aria/role-based) for targeted actions.
- Act: click, type, select, etc. using a stable ref from a snapshot.
- Navigate: open a URL, click links, go back/forward, or jump to a new page.
- Read: extract visible text or attributes from elements via refs.
- Screenshot: save a PNG/JPEG of the current view.
- Handle dialogs: accept or dismiss modal dialogs, confirm prompts, etc.
- Evaluate: run small in-page scripts to query state or extract data.
- Console/log: view basic console output or errors from the page (where supported). "
way that lets me perform actions, so I can’t execute the click with a guaranteed result." or "I can’t click it through the browser control right now—the bridge is flaky and isn’t reliably returning refs."
Then I asked at least you can scroll down to the bottom? OpenClaw says: "I can’t perform the scroll via the browser control right now due to the bridge/tool limitations."
Then I asked "so what are your capabilities", he said " - Snapshot: capture the page state and produce stable element refs (aria/role-based) for targeted actions.
- Act: click, type, select, etc. using a stable ref from a snapshot.
- Navigate: open a URL, click links, go back/forward, or jump to a new page.
- Read: extract visible text or attributes from elements via refs.
- Screenshot: save a PNG/JPEG of the current view.
- Handle dialogs: accept or dismiss modal dialogs, confirm prompts, etc.
- Evaluate: run small in-page scripts to query state or extract data.
- Console/log: view basic console output or errors from the page (where supported). "
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
83,815Members
Resources
Similar Threads
Was this page helpful?
