Browser Operator

Choose read mode for research and extraction. Choose write mode only when the team must submit, update, or click something consequential. Add only the domains needed for the task.

1. 1Add a Browser Operator tool to the responsible card.
2. 2Set read or write mode and approve the domain list.
3. 3Run the team and watch the resizable browser dock on the right.

Codelit pauses before sensitive actions and shows the proposed operation. Approve, reject, or take over the live session. A takeover stops autonomous interaction until you hand control back.

Successful and failed sessions store bounded audit events, final URLs, DOM evidence, and screenshots when available. Saved browser contexts are encrypted and can be disabled or deleted in Settings.

Place a verifier after Browser Operator so the team separates visible evidence from inference. Add a typed connected-app action after that verifier to post the result to Slack, update GitHub, create a Jira or Linear issue, or append a Notion page. Codelit renders the final action from the browser-grounded handoff, pauses for approval, and keeps both the browser evidence and provider delivery proof in the Run Receipt.

1. 1Run Browser Operator inside its approved domains.
2. 2Have the next Agent finalize the report from current DOM and screenshot evidence.
3. 3Approve the exact connected-app action and verify its native receipt in Activity.

Choose Teach by example on a Browser Operator capability when it is easier to demonstrate a task than describe it. Codelit records semantic checkpoints such as accessible targets and navigation boundaries, replaces entered values with run-time inputs, and creates a reviewable Flow.

1. 1Enter one website and task, then approve the temporary browser session.
2. 2Perform one meaningful action at a time and save the step.
3. 3Review protected or uncertain steps, then add the automatically checked draft to Flow.