Document steering and the follow-up queue

.cspell.json

@@ -29,6 +29,7 @@
     "Cloudflare",
     "Cloudflared",
     "Composio",
+    "Cowork",
     "Customization",
     "DCR",
     "Disconnect",

features/chat.mdx

@@ -52,6 +52,52 @@ Work with files in your workspace. Read a report, write a markdown file, append
 
 Use any connected app as part of an answer. Pull the latest message from Sarah, search for a Slack thread, list open PRs that need review.
 
+## While Fluso is working
+
+The composer stays live during a run. A message you send mid-run takes one of two paths: it can **steer** the work in progress, or wait in a queue and run right after.
+
+| | Steer | Queued follow-up |
+|---|---|---|
+| Send with | **Enter** | **Tab**, or **Queue follow-up** in the menu |
+| When it runs | Inside the current run, at its next pause | After the run finishes, as its own request |
+| Use it for | Corrections to the work being made now | The next task |
+
+### Steering
+
+Start typing while Fluso works and the send button switches to **Steer now**. Press Enter, and the message folds into the running reply at its next pause between steps. Nothing restarts: work already done stands, and the same reply answers both.
+
+<Frame>
+  <img src="/images/features/chat/steer-send-button.png" alt="The chat composer in dark mode with a correction typed while the agent is running. The send button shows a Steer now tooltip, with a small arrow button next to it." />
+</Frame>
+
+Steer with whatever you would say to a colleague mid-task: *"one page max"*, *"skip Fireflies"*, *"use last quarter's numbers"*. If you can see a reply going wrong, steer it instead of waiting for a wrong answer to finish. And if the run happens to end just before your message arrives, the message starts a reply of its own — either way it runs once.
+
+### Queuing follow-ups
+
+When the message is the next task rather than a correction, press **Tab** instead, or pick **Queue follow-up** from the small arrow next to the send button.
+
+<Frame>
+  <img src="/images/features/chat/active-run-menu.png" alt="The active-run menu open next to the send button, with two items: Queue follow-up with a Tab shortcut, and Stop agent." />
+</Frame>
+
+Queued follow-ups wait above the composer and run in order once the current run finishes, each as its own request, exactly once. The queue lives on the server: reload the app or switch devices and it is still there. Attachments queue along with the text.
+
+<Frame>
+  <img src="/images/features/chat/queued-tray.png" alt="Two queued follow-ups stacked above the composer, each row with edit, send now, and cancel buttons. The send button below shows the square stop control of an active run." />
+</Frame>
+
+Each row has three controls. **Edit** (the pencil) moves the message into the composer; press Enter to save, and what runs is the last version you saved. **Send now** (the play button) skips the wait: the follow-up folds into the live run as a steer, or starts immediately if nothing is running. **Cancel** (the cross) removes it for good.
+
+Two safeguards worth knowing. If a follow-up starts while you are still editing it, the save is refused with a notice and your text stays in the composer — nothing sends twice. And if a follow-up cannot reach the server, the row is marked failed with a retry, not silently dropped.
+
+### Stopping pauses the queue
+
+Stop the run — the square button, or **Stop agent** in the same menu — and queued follow-ups pause instead of firing into a conversation you cut short. The same happens if the run fails. Paused follow-ups never start on their own: **Send now** runs one, **Edit** rewrites it first, **Cancel** clears it.
+
+<Frame>
+  <img src="/images/features/chat/tray-paused.png" alt="Two queued follow-ups marked with a Paused badge after the run was stopped." />
+</Frame>
+
 ## A few habits
 
 - One topic per thread. Switching subjects mid-thread blurs the context Fluso uses. The cost of starting a new thread is zero.

going-deeper.mdx

@@ -18,7 +18,7 @@ A few patterns worth practising.
 
 **Outcomes, not steps.** Don't tell Fluso *how* to do something. "Search Gmail for messages from Sarah, read them, then tell me what she wants" is what a robot would type. "What did Sarah email me about?" is what you'd actually say. Both work. The second is faster.
 
-**Steer mid-stream.** Replies stream token by token. If you can see one going wrong, type into the box: *"actually, just today"*, *"shorter"*, *"include the attachments too"*. Don't wait for it to finish a wrong answer.
+**Steer mid-stream.** Replies stream token by token. If you can see one going wrong, type into the box: *"actually, just today"*, *"shorter"*, *"include the attachments too"*. Don't wait for it to finish a wrong answer. The full mid-run controls — steering and the editable follow-up queue — are in [While Fluso is working](/features/chat#while-fluso-is-working).
 
 <Note>
 🎥 **GIF — 5 to 8 seconds. P2.**

quickstart.mdx

@@ -153,6 +153,8 @@ If the answer is too broad, ask for a narrower version:
 
 > *"Turn this into a task list I can review today."*
 
+You do not have to wait for a run to finish before redirecting it. A message sent mid-run steers the work in progress — see [While Fluso is working](/features/chat#while-fluso-is-working).
+
 ## Step 7: make one habit
 
 Once Fluso gives you one useful result, make that workflow repeatable.