What is your best OpenClaw controlling a browser experience and setup?
Some of my workflows require that I give the lobster access to some websites, unfortunately they do not have an API and I am a bit paranoid with passwords, and some of them also have 2FA/MFA so I need to attach a tab via the browser relay. The connection appears to be very flimsy in Molty's own words " The Browser Relay attachment is stateful and fragile. Across restarts/new sessions, the relay/CDP permissions/target IDs can change."
So I am thinking - I would be hardly the only one having issues with this.
What I've tried :
- using the headless playwright approach, but this is flagging bot detections
- restarting the session of the browser relay ( this actually works sometimes, but I can't pin why it's not working constantly )
- making Molty do a full fledged investigation on the issue and this is the result :
What’s happening (in plain terms):
• I can see the tab via the Chrome relay (browser.tabs shows it).
• But when I try to actually control it (snapshot/click/type), Chrome/CDP refuses with:
Target.attachToBrowserTarget: Not allowed
• That error is a hard block at the Chrome debugging protocol level. It’s not “I didn’t try enough”; it’s Chrome saying “no, you can’t attach to this target”.
Why it can work in one session and break in the next:
• Sometimes you end up in a state where the tab is listed but CDP attachment is disallowed (common when the relay isn’t actually attached, or Chrome/extension lost the attachment handshake).
• Also possible: tab is in a special process/isolation state that blocks attach until a fresh tab is opened after attaching.
My setup is a Molty on an old laptop running in Ubuntu ; (Clawdbot) version: 2026.1.24-3
So is there a solid setup that allows the lobster to control a browser, navigate and issue commands reliably?
Browser Use comes to mind, but I will be very glad to see someone else's setup and ideas on this one
So I am thinking - I would be hardly the only one having issues with this.
What I've tried :
- using the headless playwright approach, but this is flagging bot detections
- restarting the session of the browser relay ( this actually works sometimes, but I can't pin why it's not working constantly )
- making Molty do a full fledged investigation on the issue and this is the result :
What’s happening (in plain terms):
• I can see the tab via the Chrome relay (browser.tabs shows it).
• But when I try to actually control it (snapshot/click/type), Chrome/CDP refuses with:
Target.attachToBrowserTarget: Not allowed
• That error is a hard block at the Chrome debugging protocol level. It’s not “I didn’t try enough”; it’s Chrome saying “no, you can’t attach to this target”.
Why it can work in one session and break in the next:
• Sometimes you end up in a state where the tab is listed but CDP attachment is disallowed (common when the relay isn’t actually attached, or Chrome/extension lost the attachment handshake).
• Also possible: tab is in a special process/isolation state that blocks attach until a fresh tab is opened after attaching.
My setup is a Molty on an old laptop running in Ubuntu ; (Clawdbot) version: 2026.1.24-3
So is there a solid setup that allows the lobster to control a browser, navigate and issue commands reliably?
Browser Use comes to mind, but I will be very glad to see someone else's setup and ideas on this one
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
134,827Members
Resources
Similar Threads
Was this page helpful?
