Permission prompts inline

Where a tool call needs your approval before it can run, that request appears directly inline in the transcript at the exact point it comes up, not gathered into a separate approval screen shown before or after the transcript.

This keeps the prompt in context: you see everything that happened before it, and everything that happens after your answer continues in the same place, rather than jumping between a transcript view and a separate approvals view.

The full mechanics of what these prompts show and how each possible answer is handled are covered in Permissions and safety.

Everything that happened before the prompt stays visible above it while you decide; you are never asked to approve something without the surrounding context of how the turn arrived at that point still on screen.

Answering the prompt resumes the transcript exactly where it left off, the next tool header or the final response appearing immediately after your answer, with no separate loading step or context switch back into the conversation.