4.9.0 — The prompts pane
A snapshot as of 4.9.0 (released 2026-08-17). It will go stale as the app moves on — that is expected, and the living reference is the guide each section links to.
npx mulmoterminal@latest
Nothing to configure. It is on as soon as you update. This page is about where to find the new pane and what it reads.
No screenshots on this page yet. The prompts pane is visible enough to want one, and it is deliberately deferred here rather than shipped wrong — capturing it means pointing
HOMEat a scratch directory so real project paths do not leak into the image. Until then the description below is precise about where the button is.
The prompts you sent, as a list
The problem. With several cells running, it stops being obvious which one you told what. The clues that exist each answer with exactly one item:
| Where | What it shows |
|---|---|
| The cell header’s task line | the latest prompt, one line |
| The Activity timeline | what the agent ran — its tools, not your instructions |
| The roster row | what the agent said |
None of them is a list of what you asked for.
Where it is. Enlarge a cell, then press the forum button in the cell’s header (its tooltip reads “Show the prompts you sent this session”). The pane opens on the right, beside the terminal. Press it again to close.
The button appears only while the cell is enlarged, and it appears on every cell type — a cell with no agent opens a pane that says so, which is easier to understand than a button that is missing for a reason you cannot see.
What it does.
- Newest first.
- Read-only. No jumping to that point in the conversation, no re-sending. Click a long prompt and it expands in place.
- Leave it open. It keeps up on its own as you send more.
- Claude and Codex. Other agents show an empty pane.
What it reads, and why that matters. It reads what you typed — for Claude that is ~/.claude/history.jsonl, not the transcript. That is a measured distinction, not a preference:
- A prompt you send while a turn is already running never appears in the transcript as a user record. It lands as
queue-operation/attachment. Those interruptions are exactly the instructions you are most likely to forget giving. - Text that a skill injects does appear in the transcript as a user record, and you never typed it.
So a pane built on the transcript would drop the half you need and add a half you did not write.
After /clear. The pane draws a line there. Prompts from before the clear are not shown — the same rule the header, the AI title and the latest reply already follow. That boundary survives a server restart.
After /compact. Nothing changes; the pane keeps showing the whole session. (/compact does not reissue the session id — measured across 95 compacted transcripts.)
→ Reference: Features
Fixes you might have hit
- The dev server restarted forever when the port was taken. A port already in use made it exit, which made it restart, which made it exit again. It reports the collision and stops now. If you saw an endless restart loop in the console, this was it.
- A recorded timestamp is no longer editable from the collection pane. It is the server’s record of when a row was written, not a field to fill in.
How to tell you have it
npx mulmoterminal@latest --version
4.9.0. Then enlarge any Claude or Codex cell — the forum button is in the header, and the pane it opens lists the prompts you have sent that session.
Full detail on every change: Changelog.