Explain
A separate Q&A window next to the session. Ask about what the agent is doing without cutting into the main conversation or spending the agent's context.
How to use it
- Click the Explain button next to Verbose or press
Ctrl+Shift+E. The floating window can be dragged anywhere. - Select a passage in the transcript to attach it as a quote, then type your question.
- Answers include a button that jumps to the exact turn in the chat and list any extra files that were read.
- If you want the main agent to act on it, click Send to main composer; the content is copied to the composer so you can edit it before sending.
Limits
- It only reads the transcript and, when needed, files in the session using Read, Grep and Glob. It doesn't run skills, edit files, or send anything to the main agent.
- It uses its own model, which you can change in the window. The top bar shows how many turns have been read and how much context is in use.
- History is private by default. If you turn on sharing, people viewing the session can read the history but can't ask questions.