diff --git a/.claude-plugin/plugin.json b/.claude-plugin/plugin.json
index 354e9c9c..6cff9a7b 100644
--- a/.claude-plugin/plugin.json
+++ b/.claude-plugin/plugin.json
@@ -22,7 +22,8 @@
           {
             "type": "command",
             "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/hook-handler.js\" PreToolUse",
-            "timeout": 5
+            "timeout": 5,
+            "async": true
           }
         ]
       }
@@ -34,7 +35,8 @@
           {
             "type": "command",
             "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/hook-handler.js\" PostToolUse",
-            "timeout": 5
+            "timeout": 5,
+            "async": true
           }
         ]
       }
diff --git a/client/src/components/Sagi/PetSprite.tsx b/client/src/components/Sagi/PetSprite.tsx
index 95b1c5f1..38e74ce4 100644
--- a/client/src/components/Sagi/PetSprite.tsx
+++ b/client/src/components/Sagi/PetSprite.tsx
@@ -58,6 +58,8 @@ const ROW: Record<Mood, number> = {
 /** What the user is doing to the pet, which outranks its mood. */
 export type PetAction = "run-left" | "run-right" | "wave";
 const ACTION_ROW: Record<PetAction, number> = { "run-right": 1, "run-left": 2, wave: 3 };
+/** Loops of a row before the sprite rests on its first frame. */
+const SPRITE_LOOPS = 4;
 
 export function PetSprite({
   id,
@@ -95,13 +97,17 @@ export function PetSprite({
     const own = play(row, sleeping ? SLEEP_SLOWDOWN : 1);
     let list = MOMENTS.has(row) ? [...own, ...own, ...own] : own;
     let i = 0;
+    let loops = 0;
     let timer = 0;
-    // `i` always indexes `list`: it wraps at its end.
+    // `i` always indexes `list`: it wraps at its end. A few loops, then the
+    // first frame rests — a sprite that never stops re-paints the page several
+    // times a second for as long as it is open; a new mood or action plays again.
     const tick = () => {
       timer = window.setTimeout(() => {
         if (++i === list.length) {
           i = 0;
           if (MOMENTS.has(row)) list = play(0);
+          if (++loops >= SPRITE_LOOPS) return setCell(list[0]!);
         }
         setCell(list[i]!);
         tick();
diff --git a/client/src/components/Sagi/sagi.css b/client/src/components/Sagi/sagi.css
index a728f7c1..36609679 100644
--- a/client/src/components/Sagi/sagi.css
+++ b/client/src/components/Sagi/sagi.css
@@ -1,8 +1,12 @@
 /**
  * sagi.css - animations + mood expressions for the Sagi companion.
  * Colors mirror the app's theme tokens (accent #6366f1/#818cf8, surface scale)
- * with warm pink accents (ears, cheeks, nose) for cuteness. All continuous
- * motion is disabled under [data-reduced="1"] and prefers-reduced-motion.
+ * with warm pink accents (ears, cheeks, nose) for cuteness. No mood animates
+ * forever: each plays a few cycles when the mood begins and then rests — an
+ * infinite animation makes the browser draw a new frame 60 times a second for
+ * as long as the page is open (the drop-shadow filter re-rendered each time),
+ * which made the whole desktop stutter. All motion is disabled under
+ * [data-reduced="1"] and prefers-reduced-motion.
  * Transform origins are the delicate part - see the comments on .sagi-avatar
  * and .sagi-eye-blink for which reference box each one resolves against.
  * @author Nguyễn Ngọc Trí Vĩ <vinnt@smartgift.vn>
@@ -206,20 +210,20 @@
 
 /* idle: gentle breathing + occasional blink */
 .sagi-avatar[data-mood="idle"] {
-  animation: sagi-breathe 4s ease-in-out infinite;
+  animation: sagi-breathe 4s ease-in-out 0s 3;
 }
 /* Blink lives on the inner group so it never overrides the outer group's
    eye-tracking translate (which would freeze the eyes - see the CSS-blink
    note in SagiAvatar.tsx). */
 .sagi-avatar[data-mood="idle"] .sagi-eye-blink {
-  animation: sagi-blink 5s ease-in-out infinite;
+  animation: sagi-blink 5s ease-in-out 0s 3;
 }
 /* happy: whole-mascot bob + twinkling sparkle */
 .sagi-avatar[data-mood="happy"] {
   animation: sagi-bob 0.5s ease-in-out 0s 4;
 }
 .sagi-avatar[data-mood="happy"] .sagi-sparkle {
-  animation: sagi-sparkle-twinkle 0.9s ease-in-out infinite;
+  animation: sagi-sparkle-twinkle 0.9s ease-in-out 0s 4;
   transform-origin: 797px 274px; /* sparkle centre, mascot 1050,380 */
 }
 /* worried: alarmed shake */
@@ -228,19 +232,19 @@
 }
 /* stuck: slow, drooping breathe (fatigued) */
 .sagi-avatar[data-mood="stuck"] {
-  animation: sagi-droop 2.4s ease-in-out infinite;
+  animation: sagi-droop 2.4s ease-in-out 0s 3;
 }
 /* thinking: rocking head tilt */
 .sagi-avatar[data-mood="thinking"] {
-  animation: sagi-rock 1.8s ease-in-out infinite;
+  animation: sagi-rock 1.8s ease-in-out 0s 4;
 }
 /* watching: attentive sway while a session is live */
 .sagi-avatar[data-mood="watching"] {
-  animation: sagi-sway 3.2s ease-in-out infinite;
+  animation: sagi-sway 3.2s ease-in-out 0s 3;
 }
 /* sleeping: slow breathe, droop */
 .sagi-avatar[data-mood="sleeping"] {
-  animation: sagi-breathe 5s ease-in-out infinite;
+  animation: sagi-breathe 5s ease-in-out 0s 3;
 }
 
 /* ── reduced motion: kill all continuous animation ── */
@@ -339,7 +343,7 @@
   align-items: center;
   justify-content: center;
   border: 2px solid #0c0c14;
-  animation: sagi-error-pulse 1.2s ease-in-out infinite;
+  animation: sagi-error-pulse 1.2s ease-in-out 0s 5;
 }
 .sagi-bubble {
   max-width: 16rem;
@@ -375,7 +379,7 @@
 /* An uploaded still image has no frames of its own: it bobs, faster while an
    agent works and slower asleep. A Codex spritesheet animates by its frames. */
 .sagi-pet-image {
-  animation: sagi-pet-bob 2.6s ease-in-out infinite;
+  animation: sagi-pet-bob 2.6s ease-in-out 0s 3;
   -webkit-user-drag: none;
 }
 .sagi-pet-image[data-mood="thinking"] {
diff --git a/client/src/components/Sagi/useSagiBrain.ts b/client/src/components/Sagi/useSagiBrain.ts
index e34ac9bd..37ff132f 100644
--- a/client/src/components/Sagi/useSagiBrain.ts
+++ b/client/src/components/Sagi/useSagiBrain.ts
@@ -77,6 +77,17 @@ import {
 import { pickQuip } from "./quips";
 import { sagiPrefs } from "./prefs";
 
+/** `next` differs from `prev` only by a lastActivityAt less than a second
+ *  newer — nothing the mascot shows, at the moods' 1 s resolution. */
+function onlyActivityMoved(prev: SagiState, next: SagiState): boolean {
+  if (next === prev) return true;
+  if (next.lastActivityAt - prev.lastActivityAt >= 1000) return false;
+  for (const k of Object.keys(next) as (keyof SagiState)[]) {
+    if (k !== "lastActivityAt" && next[k] !== prev[k]) return false;
+  }
+  return true;
+}
+
 const BUBBLE_MS = 4500;
 // Minimum gap between non-error bubbles, so a burst of activity doesn't spam.
 const BUBBLE_THROTTLE_MS = 3000;
@@ -153,7 +164,10 @@ export function useSagiBrain(): SagiBrain {
       setState((prev) => {
         const { state: next, pulse } = reduceSagi(prev, msg, t);
         if (pulse) showBubble(pickQuip(pulse), pulse === "error");
-        return next;
+        // Most frames only move lastActivityAt, which the moods read at a 1 s
+        // resolution: keep the old state then, so a hook burst (several
+        // frames a second) does not re-render the mascot on every frame.
+        return onlyActivityMoved(prev, next) ? prev : next;
       });
     });
     const unsubConn = eventBus.onConnection((connected) => {
@@ -169,15 +183,25 @@ export function useSagiBrain(): SagiBrain {
   // happy/worried) re-evaluate without needing a new event.
   useEffect(() => {
     // Hidden tab: no re-render; the first visible tick catches the mood up.
+    // Only a tick that changes the mood re-renders: an idle mascot is not
+    // drawn again every second for nothing.
     const id = setInterval(() => {
-      if (document.visibilityState === "visible") setTick(Date.now());
+      if (document.visibilityState !== "visible") return;
+      const now = Date.now();
+      if (deriveMood(stateRef.current, now) !== moodRef.current) setTick(now);
     }, 1000);
     return () => clearInterval(id);
   }, []);
 
   useEffect(() => () => clearTimeout(bubbleTimer.current), []);
 
-  const mood = useMemo(() => deriveMood(state, tick), [state, tick]);
+  // Judged at render time: `tick` only moves when the mood would change, so
+  // it may be old when a frame re-renders.
+  const mood = useMemo(() => deriveMood(state, Math.max(tick, Date.now())), [state, tick]);
+  const stateRef = useRef(state);
+  stateRef.current = state;
+  const moodRef = useRef(mood);
+  moodRef.current = mood;
   const status = useMemo(() => statusOf(state), [state]);
 
   const dismissBubble = useCallback(() => {
diff --git a/client/src/components/conversation/ConversationView.tsx b/client/src/components/conversation/ConversationView.tsx
index 81c6cdf6..2df66a35 100644
--- a/client/src/components/conversation/ConversationView.tsx
+++ b/client/src/components/conversation/ConversationView.tsx
@@ -61,7 +61,7 @@
  *
  * ----------------------------------------------------------------------------- */
 
-import { useEffect, useState, useCallback, useMemo, useRef } from "react";
+import { memo, useEffect, useState, useCallback, useMemo, useRef } from "react";
 import { Loader2, ArrowDown, MessagesSquare, RefreshCw } from "lucide-react";
 import { api } from "../../lib/api";
 import { eventBus } from "../../lib/eventBus";
@@ -112,7 +112,7 @@ interface ConversationViewProps {
   onNotStarted?: (notStarted: boolean) => void;
 }
 
-export function ConversationView({
+function ConversationViewPane({
   sessionId,
   initialTranscriptId,
   fill = false,
@@ -646,3 +646,7 @@ export function ConversationView({
     </div>
   );
 }
+
+/** Memoized: a chat's prompt box lives in its parent (PaneChat), and every
+ *  keystroke there re-rendered the whole conversation. */
+export const ConversationView = memo(ConversationViewPane);
diff --git a/client/src/components/conversation/MessageList.tsx b/client/src/components/conversation/MessageList.tsx
index 8919ffde..813802e5 100644
--- a/client/src/components/conversation/MessageList.tsx
+++ b/client/src/components/conversation/MessageList.tsx
@@ -633,7 +633,7 @@ const EditedFiles = memo(
     sameResults(n.blocks, p.toolResultMap, n.toolResultMap)
 );
 
-export function MessageList({
+function MessageListView({
   messages,
   loading,
   onOpenFile,
@@ -1550,3 +1550,6 @@ const MessageRow = memo(
   MessageRowView,
   (p, n) => p.msg === n.msg && sameResults(n.msg.content, p.toolResultMap, n.toolResultMap)
 );
+
+/** Memoized: its owner re-renders on scroll state and its parent's typing. */
+export const MessageList = memo(MessageListView);
diff --git a/client/src/components/run/PaneChat.tsx b/client/src/components/run/PaneChat.tsx
index b10f7062..aa638b67 100644
--- a/client/src/components/run/PaneChat.tsx
+++ b/client/src/components/run/PaneChat.tsx
@@ -467,6 +467,19 @@ export const PaneChat = memo(function PaneChat({
     []
   );
 
+  // Stable, so typing in the prompt box does not re-render the conversation
+  // (ConversationView is memoized).
+  const openTerminal = useCallback(() => showTerminal(runId, sessionId), [runId, sessionId]);
+  const openFile = useCallback(
+    (path: string) => {
+      // The workspace's Changes panel when the file is in its folder;
+      // else this view's own preview.
+      if (laneId !== null && openInChanges(laneId, cwd, path)) return;
+      setPreview(path);
+    },
+    [laneId, cwd]
+  );
+
   // Before the session starts the CLI may be asking in the terminal (trust
   // this folder? with "No, exit" preselected): an Enter from here answers it.
   // Asked, not refused: whether the session row appears right after the
@@ -515,15 +528,10 @@ export const PaneChat = memo(function PaneChat({
             follow={follow}
             waiting={asking}
             stopped={stops}
-            onOpenTerminal={() => showTerminal(runId, sessionId)}
+            onOpenTerminal={openTerminal}
             onContext={onContext}
             onNotStarted={setNotStarted}
-            onOpenFile={(path) => {
-              // The workspace's Changes panel when the file is in its folder;
-              // else this view's own preview.
-              if (laneId !== null && openInChanges(laneId, cwd, path)) return;
-              setPreview(path);
-            }}
+            onOpenFile={openFile}
           />
           {!outlineOpen && !preview && (
             <button
diff --git a/client/src/components/run/TerminalView.tsx b/client/src/components/run/TerminalView.tsx
index 33d99d18..fffc5244 100644
--- a/client/src/components/run/TerminalView.tsx
+++ b/client/src/components/run/TerminalView.tsx
@@ -62,7 +62,7 @@ import { useTheme } from "../../hooks/useTheme";
 import { terminalThemeFor } from "../../lib/terminalTheme";
 import { api } from "../../lib/api";
 import { writeClipboard } from "../../lib/clipboard";
-import { notePaneFocus, registerPane } from "../../lib/panePaste";
+import { MENU_CURSOR_ROW, notePaneFocus, registerPane } from "../../lib/panePaste";
 import { attachImeInput } from "../../lib/imeInput";
 import { TERMINAL_FONTS, readTerminalFont } from "../../lib/terminalFont";
 
@@ -649,16 +649,27 @@ export const TerminalView = memo(function TerminalView({
           if (wsRef.current?.readyState === WebSocket.OPEN) wsRef.current.send(data);
         },
         // The rows on screen, dim cells (a TUI's placeholder) left out.
-        () => {
+        // `menuOnly` (the chat's 700 ms menu poll): plain text first, and the
+        // cell-by-cell read only when a row could be a menu's cursor row.
+        (menuOnly = false) => {
           const term = termRef.current;
           if (!term) return null;
           const buf = term.buffer.active;
+          if (menuOnly) {
+            let maybe = false;
+            for (let y = 0; y < term.rows && !maybe; y++)
+              maybe = MENU_CURSOR_ROW.test(
+                buf.getLine(buf.baseY + y)?.translateToString(true) ?? ""
+              );
+            if (!maybe) return null;
+          }
           const rows: string[] = [];
+          const reuse = buf.getNullCell();
           for (let y = 0; y < term.rows; y++) {
             const line = buf.getLine(buf.baseY + y);
             let row = "";
             for (let x = 0; line && x < line.length; x++) {
-              const cell = line.getCell(x);
+              const cell = line.getCell(x, reuse);
               // Width 0: the right half of a wide character.
               if (cell && cell.getWidth() && !cell.isDim()) row += cell.getChars() || " ";
             }
diff --git a/client/src/index.css b/client/src/index.css
index fbcb7a61..8db328b3 100644
--- a/client/src/index.css
+++ b/client/src/index.css
@@ -530,12 +530,18 @@
   }
 }
 
+/* A few pulses when the dot appears, then it rests lit. Status dots sit in
+   always-visible chrome; an infinite animation there makes the browser draw a
+   frame 60 times a second for as long as the page is open, which made the
+   whole desktop stutter. A state change re-adds the class and pulses again. */
 .animate-pulse-dot {
-  animation: pulse-dot 2s cubic-bezier(0.4, 0, 0.6, 1) infinite;
+  animation: pulse-dot 2s cubic-bezier(0.4, 0, 0.6, 1) 3;
 }
 
-/* The chat's live step, like Claude Code's: a light sweeping over the words.
-   Reduced motion stops it on the muted colour. */
+/* The chat's live step, like Claude Code's: a light sweeping over the words —
+   a few sweeps, then it rests on the muted colour (a background-position
+   animation repaints the text every frame; the spinner beside it keeps saying
+   the step runs). Reduced motion stops it at once. */
 .text-shimmer {
   background: linear-gradient(
       90deg,
@@ -547,7 +553,7 @@
   -webkit-background-clip: text;
   background-clip: text;
   color: transparent;
-  animation: text-shimmer 2.5s linear infinite;
+  animation: text-shimmer 2.5s linear 3;
 }
 
 @keyframes text-shimmer {
diff --git a/client/src/lib/__tests__/panePaste.test.ts b/client/src/lib/__tests__/panePaste.test.ts
index 20a383d5..e2495762 100644
--- a/client/src/lib/__tests__/panePaste.test.ts
+++ b/client/src/lib/__tests__/panePaste.test.ts
@@ -24,6 +24,7 @@ import {
   readRunMenu,
   registerPane,
   takeRunInput,
+  MENU_CURSOR_ROW,
 } from "../panePaste";
 
 const RULE = "─".repeat(60);
@@ -311,3 +312,17 @@ const CODEX_APPROVAL = [
   // "Press enter to confirm or esc to cancel", its dim cells left out.
   "  enteresc",
 ];
+
+describe("MENU_CURSOR_ROW", () => {
+  // The chat's 700 ms poll reads the screen cell by cell only when a row
+  // matches it, so every menu parsePaneMenu finds must have such a row.
+  it("matches a row of every menu, and no row of an idle prompt", () => {
+    const has = (rows: string[]) => rows.some((r) => MENU_CURSOR_ROW.test(r));
+    for (const menu of [ASK, MULTI, PERMISSION, CODEX_APPROVAL]) {
+      expect(parsePaneMenu(menu)).not.toBeNull();
+      expect(has(menu)).toBe(true);
+    }
+    expect(has(CLAUDE)).toBe(false);
+    expect(has(CODEX)).toBe(false);
+  });
+});
diff --git a/client/src/lib/panePaste.ts b/client/src/lib/panePaste.ts
index a0124cb3..39c81249 100644
--- a/client/src/lib/panePaste.ts
+++ b/client/src/lib/panePaste.ts
@@ -27,7 +27,8 @@ interface Pane {
   paste: (text: string, submit?: boolean) => void;
   visible: () => boolean;
   key?: (data: string) => void;
-  screen?: () => PaneScreen | null;
+  /** `menuOnly`: null at once unless a row could be a menu's cursor row. */
+  screen?: (menuOnly?: boolean) => PaneScreen | null;
 }
 const panes = new Map<string, Pane>();
 let lastFocused: string | null = null;
@@ -39,7 +40,7 @@ export function registerPane(
   paste: (text: string, submit?: boolean) => void,
   visible: () => boolean = () => true,
   key?: (data: string) => void,
-  screen?: () => PaneScreen | null
+  screen?: (menuOnly?: boolean) => PaneScreen | null
 ): () => void {
   const entry = { paste, visible, key, screen };
   panes.set(runId, entry);
@@ -175,6 +176,8 @@ export type PaneMenu = {
 };
 
 const OPTION = /^ {0,4}([❯›][ \u00a0]?)? *(\d{1,2})\. +(.*)$/;
+/** A row a menu's cursor sits on (a numbered option or Submit): a menu has one. */
+export const MENU_CURSOR_ROW = /^ *[❯›][ \u00a0]* *(?:\d{1,2}\. |Submit\s*$)/;
 const SUBMIT = /^ *([❯›])?[ \u00a0]*Submit$/;
 const RULE_ROW = /^ *[─╌]{3,} *$/;
 const CHECK = /^\[([ ✔])\] *(.*)$/;
@@ -251,7 +254,9 @@ function typeRow(options: PaneMenuOption[]): number {
 /** The menu `runId`'s visible pane shows now, or null. */
 export function readRunMenu(runId: string): PaneMenu | null {
   const p = panes.get(runId);
-  const screen = p?.visible() && p.key ? p.screen?.() : null;
+  // Polled every 700 ms while a chat shows: a screen with no cursor row is
+  // answered from plain text, without reading it cell by cell.
+  const screen = p?.visible() && p.key ? p.screen?.(true) : null;
   return screen ? parsePaneMenu(screen.rows) : null;
 }
 
diff --git a/client/src/pages/Dashboard.tsx b/client/src/pages/Dashboard.tsx
index 7779927a..1157b20f 100644
--- a/client/src/pages/Dashboard.tsx
+++ b/client/src/pages/Dashboard.tsx
@@ -2664,7 +2664,7 @@ function SystemHealthTab() {
 
           <Tip block raw={t("health.wsTip", { count: info.server.ws_connections })}>
             <div className="flex items-center gap-3 bg-status-success/5 px-3 py-2.5 rounded-lg border border-status-success/10 cursor-default">
-              <Activity className="w-3.5 h-3.5 text-status-success animate-pulse" />
+              <Activity className="w-3.5 h-3.5 text-status-success animate-pulse-dot" />
               <div>
                 <p className="text-[10px] text-status-success font-medium">
                   {t("health.wsActive")}
diff --git a/client/tailwind.config.js b/client/tailwind.config.js
index 45e37218..6dd601a8 100644
--- a/client/tailwind.config.js
+++ b/client/tailwind.config.js
@@ -122,7 +122,10 @@ export default {
         serif: ["ui-serif", "Cambria", "Times New Roman", "Times", "serif"],
       },
       animation: {
-        "pulse-slow": "pulse 3s cubic-bezier(0.4, 0, 0.6, 1) infinite",
+        // Status indicators pulse a few times and rest (never `infinite`: an
+        // always-on animation keeps the browser drawing 60 frames a second).
+        "pulse-slow": "pulse 3s cubic-bezier(0.4, 0, 0.6, 1) 3",
+        ping: "ping 1s cubic-bezier(0, 0, 0.2, 1) 4",
         "fade-in": "fadeIn 0.3s ease-out",
         "slide-up": "slideUp 0.3s ease-out",
       },
diff --git a/docs/API.md b/docs/API.md
index 1f0d71b5..9ca5436a 100644
--- a/docs/API.md
+++ b/docs/API.md
@@ -2015,6 +2015,7 @@ Response `Content-Type: text/plain; version=0.0.4; charset=utf-8`.
 | `agenthub_build_info` | gauge | `version` | Always `1`; dashboard version rides on the label |
 | `agenthub_process_uptime_seconds` | gauge | — | Server process uptime |
 | `agenthub_process_resident_memory_bytes` | gauge | — | Server process RSS |
+| `agenthub_event_loop_delay_seconds` | gauge | `quantile` (`0.5`, `0.99`, `1` = max) | Event-loop delay over the last minute — how long requests, hooks and terminal echo waited (a laggy terminal shows here) |
 | `agenthub_sessions` | gauge | `status` (`active`/`completed`/`error`/`abandoned`) | Sessions by status |
 | `agenthub_agents` | gauge | `status` (`working`/`waiting`/`completed`/`error`) | Agents by status |
 | `agenthub_events_total` | counter | — | Total events recorded |
diff --git a/docs/HOOKS.md b/docs/HOOKS.md
index 1f9b2395..fe5e24a4 100644
--- a/docs/HOOKS.md
+++ b/docs/HOOKS.md
@@ -182,7 +182,7 @@ npm run install-hooks
 > (Escape hatch for running Claude Code *inside* the same container:
 > `AGENTHUB_ALLOW_CONTAINER_HOOKS=1 npm run install-hooks`.)
 
-For each of the eight hook types it adds one entry to `~/.claude/settings.json` that runs the handler with the type as its argument; the five tool/turn hooks (`PreToolUse`, `PostToolUse`, `Stop`, `SubagentStop`, `Notification`) carry `"matcher": "*"`. Entries that are not AgentHub's are left alone. There is nothing else to install — Claude Code runs `scripts/hook-handler.js` directly.
+For each of the eight hook types it adds one entry to `~/.claude/settings.json` that runs the handler with the type as its argument; the five tool/turn hooks (`PreToolUse`, `PostToolUse`, `Stop`, `SubagentStop`, `Notification`) carry `"matcher": "*"`. `PreToolUse` and `PostToolUse` also carry `"async": true` (`ASYNC_HOOKS` in `scripts/install-hooks.js`, and the same in `.claude-plugin/plugin.json`): they fire around every tool call, and Claude Code waits for a synchronous hook before going on — each a Node start of 0.1–0.2 s, so ~0.4 s added to every tool call. The handler only reports (it prints nothing and decides nothing), so nothing is lost by not waiting; the rarer hooks stay synchronous so a turn's start and end reach the dashboard in order. Entries that are not AgentHub's are left alone. There is nothing else to install — Claude Code runs `scripts/hook-handler.js` directly.
 
 ```json
 {
@@ -191,7 +191,7 @@ For each of the eight hook types it adds one entry to `~/.claude/settings.json`
       {
         "matcher": "*",
         "hooks": [
-          { "type": "command", "command": "node \"/path/to/agenthub/scripts/hook-handler.js\" PreToolUse" }
+          { "type": "command", "command": "node \"/path/to/agenthub/scripts/hook-handler.js\" PreToolUse", "async": true }
         ]
       }
     ],
diff --git a/docs/LANES.md b/docs/LANES.md
index 640ff302..6f81e8e3 100644
--- a/docs/LANES.md
+++ b/docs/LANES.md
@@ -394,7 +394,7 @@ The dashboard web UI merges workspaces and runs into a single **Workspace** page
 - **A workspace's sessions** — the chevron expands the row to the sessions the workspace OWNS and has not archived (`WorkspaceSessionList.tsx`: `GET /api/sessions?lane=<id>&archived=0`, the newest 20, re-read when the workspace's status or session moves; which rows are expanded is kept in `agenthub.sidebar.projects` → `expanded`). A live AI tab whose session has no row yet — a tab opened bare from `+`, before its first prompt — is listed at the top as **New session** straight from the run (its session id is known from the start), so a new tab shows in its workspace at once; the row becomes the real session when its first hook lands. The sidebar's live panes follow the Workspace page's own run-list poll (`LIVE_RUNS_EVENT`), since a session in slots 1–9 moves no workspace column. A workspace owns a session its pane held — `bindRunSession` stamps `sessions.lane_id` from the pane's run, first writer wins, a nested `claude -p` excluded — and lists nothing else: a `claude` typed into a terminal in the same folder, or history from before the workspace existed, belongs to no workspace and is found on the Sessions page. Sessions older than this ownership column are not guessed: only those `dashboard_runs` still named were backfilled. A session started by a tool with no hooks (Codex without its hooks, any catalog tool) is never stamped. A row's dot is the session's state (`sessionDot`: amber waiting for you, red failed, green a live pane, hollow at rest); double-click renames it (`PATCH {name}`), and its hover button **archives** it (`PATCH {archived: true}` — hidden from the workspace, nothing deleted; the home lists the archived ones folded with **Unarchive**). A filled dot marks a session that still has a live pane (a `GET /api/run` handle carrying its session id — every pane the dashboard starts carries `AGENTHUB_RUN_ID=<its tmux name>`, the hook handler forwards it as `run_id`, and `bindRunSession` in `server/routes/hooks.js` records the hook's session on that run: the first session always, afterwards — for a Claude pane — only a `SessionStart` that is not a fresh `startup` (`/clear`, `/resume`), so a `claude -p` the agent runs inside the pane never takes it over; the rule is Claude-only, and a Codex or Antigravity pane follows every new session): clicking it opens `/workspace?lane=<id>&run=<runId>`, which ATTACHES to that tmux session (`attachRequest` on `LaneConsolePane`) — it starts nothing and resumes nothing. The session opens on its own tab (the tab of its tmux slot): in the pane already showing it (focused), else in the focused pane; **Ctrl/Cmd+click** opens it in a split in the tab on screen (see **Split panes** below). Any other session opens in its workspace the way Claude Desktop opens an old one: `/workspace?lane=<id>&session=<id>` shows its conversation, read-only (`SessionReplay.tsx`, the console hidden — never unmounted — behind it), with a prompt box under it. Opening it resumes nothing; **sending** a message resumes it (`POST /api/lanes/:id/start {slot, resumeSessionId, prompt}` in the first free pane slot, on the tool the session ran on) and opens that pane — a resume re-reads the whole transcript into context, so it waits for that explicit send. The session's analysis page (`/sessions/<id>`: agents, events, cost, transcript) is the **Analyst** button in the session view's right panel, beside Files and Changes (disabled while the pane's run has reported no session yet).
 - **Row menu** — right-click a row, or its `⋯` button: Rename, the Model (AI Proxy) select, New workspace here…, Copy path, Reset worktree… (managed only) and Delete workspace…, each the call the rest of the UI already makes. Delete is record-only exactly as before (see "Removing a workspace").
 - **Right-click menus** — every item that has actions opens them on a right-click too, in one shell (`client/src/components/ContextMenu.tsx`: placed at the pointer, kept inside the viewport, closed by Escape, a click outside or an item). The buttons stay; the menu repeats what they do and nothing more. A session row — in the sidebar, the home's Recents and the archived fold (`lanes/SessionMenu.tsx`) — offers Open, Open in split (live), Rename and Archive (sidebar), Unarchive (archived), Session details, Copy session ID and Close tab (live); the sidebar's **New session** row Open, Open in split and Close tab. A console tab offers Open, Open in split, Session details, Copy session ID and Close tab; a pane header (split view) Show <session> here, Split right/down with a new terminal and Close pane. In the Changes panel a changed file (overview row or review header) offers Open, Stage/Unstage, Discard — two clicks, as on its trash button — or the conflict's Keep mine / Take theirs / Mark resolved, and Copy relative path; a Files tree entry Open and Copy relative path; a History commit View commit, Copy commit SHA and Copy commit message. The home's title opens the row menu, and a Context file or extra folder offers Copy path and Remove. Outside the workspace the same shell serves the Sessions table, the Dashboard's agent cards and recent events, the Activity feed and Workflow runs (`components/SessionRowMenu.tsx`: Open session details, Open in its workspace when it has one, Copy session ID, Copy folder path), an attached prompt image (View / Remove), the managed workspace card (Reset), every Settings list (Git credentials, AI Proxy, Pricing, alert rules, webhooks, remote sources, AI tools) and every Claude Code / Codex config list (skills, agents, commands, output styles and memory files via `FileItemMenu` in `components/config/MdItemList.tsx`; plugins, MCP servers, marketplaces, auth profiles). A delete or remove there takes the button's own confirm — a modal, or two clicks. Text areas (the terminal, a conversation, a diff) keep the browser's own menu. The shell stops the menu's clicks from reaching the card it was opened from.
-- **Chat view of a live pane** — each Claude or Codex SESSION shows either a chat or its terminal, its own choice: the chat/terminal button left of the pane's split buttons (`ChatToggle` in `client/src/components/run/PaneGrid.tsx`) switches it, and the choice is kept in this browser's localStorage by session id (`agenthub.chatUi.sessions`, the newest 200; `useSessionChat` in `client/src/lib/chatUi.ts`). **Settings → Appearance → Chat UI** (`ChatUiSettings.tsx`, saved per ACCOUNT on the server: `GET/PUT /api/me/prefs`, key `agenthub.chatUi`) is only the default — written onto a session the first time it is shown, so changing it later leaves every session already seen as it is. The chat needs the session id, so a pane shows it only once its session is known (`client/src/components/run/LaneConsolePane.tsx`). Chat (`client/src/components/run/PaneChat.tsx`) shows the session's conversation the way Claude Desktop does — the shared `ConversationView` in its `fill` layout: one centred column, your messages as bubbles, the reply as plain prose, every run of tool calls folded into one line named by what it did, in Claude Desktop's fixed order ("Ran 3 commands, edited a file, used a tool ›", `workLabel`) — a thinking block that carries text is the model's narration (Claude Code ≥ 2.1.280 stores real thinking empty) and reads as prose between those lines, as Claude Desktop shows it — that opens onto a framed list of its steps — each titled by its own description, the call in a capped scroll box, its output one click further — and a round ↓ back to the end — above a prompt box that grows with its text up to 40% of the screen, then scrolls. The conversation stays at its end while you are there: a new line, the running step, an image loading or the pane shown again after being hidden keeps the newest in sight (a `ResizeObserver` in `ConversationView`), until you scroll up; sending a message takes it back to the end. It lies OVER the terminal, which stays mounted and attached (the console never unmounts a terminal); a message is sent through that terminal as a bracketed paste and Enter (`submitToRun` in `client/src/lib/panePaste.ts`), exactly as if typed, so no new route or server write exists for it. On a Windows dashboard (psmux) the attach (`server/lib/pty-attach.js`) holds any input that follows a paste until 700 ms after it (`PASTE_HOLD_MS`): psmux delays a paste ~350 ms and wrote an Enter sent meanwhile INSIDE it (psmux#766), so the prompt sat in Claude's composer as `[Pasted text #1]` instead of being sent. While the session's main agent is `working` (its `UserPromptSubmit` hook seen, no `Stop` yet — `GET /api/agents?session_id=&status=working`, re-read on `agent_updated`), an empty box shows **Stop** in Send's place, like Claude Desktop: it presses Esc in the pane (`interruptRun` in the same file), the key `claude` and `codex` stop a turn with; typing a message brings Send back (Claude Code queues it). Stop also ends the conversation's live line (`stopped` on `ConversationView`/`MessageList`) until the transcript gains a newer line: an Esc before the model wrote anything puts the prompt back in Claude Code's input box and writes no `[Request interrupted by user]` line, so the transcript still ends on the prompt and would read as **Thinking…** for 15 minutes (issue #74). An Esc pressed in the terminal itself is not seen by the chat. A Codex pane without AgentHub's hooks never shows it. An image attached in the box reaches the transcript as its saved path; the user's turn shows the image itself (`GET /api/run/drop/:name`) — any other attached file as a chip (name · EXT) that opens its preview like an output — and opens it full size over the whole page (`ImageLightbox`, rendered into `document.body`, so nothing stacked over a pane covers it). The box carries Claude Desktop's controls: a `+` menu (`ChatAddMenu.tsx`, also in the workspace home's prompt box) — Upload or add files; what already exists, attached like an upload (a chip, its path sent with the prompt): **Project files** (a search of the box's folder, folders too — `GET /api/run/files?dirs=1` — sent as `@path`), **Context files** (the workspace's own, `GET /api/lanes/:id/sources`, sent as `@<stored path>`) and **Recent uploads** (files uploaded through a prompt box in the last day, kept in this browser's localStorage `agenthub.prompt.recentUploads` — the server prunes an upload after a day); a file or folder dragged onto anywhere in a prompt box (not only its text area) attaches the same way — from the Changes panel's Files tree, its search, a Changes-list row, or a workspace's Context files and Folder list; a row from an extra folder's section (`data-drag-root`) and a context file or folder carry their absolute path, since they lie outside the pane's folder (`client/src/lib/promptAttach.ts`); Skills (yours and the project's), Plugins → a plugin → the skills and commands it ships, each put at the front of the message as `/<name>` / `/<plugin>:<name>`, and Manage links to the Claude config page — and, for a Claude pane, the Model/Effort chip, a change typed into the pane as `/model <id>` / `/effort <level>`, and the permission mode beside `+` in the box. A menu opened in a split pane (a `Select`, the Model chip) flips up or down by the room inside ITS pane (`data-menu-bounds` on the pane, `menuRoom` in `client/src/components/Select.tsx`), never the window's — a menu running past the pane's edge is painted over by the pane below. A running session changes mode only by Shift+Tab, so the server steps it and reads each step back off the pane (`POST /api/run/:id/permission-mode`, `server/lib/pane-mode.js`): it lands on exactly the mode picked, presses nothing while a menu hides the mode line, and returns to where it began when the mode is not in that session's cycle (Bypass only when the session was started with it). The mode belongs to that session only. A menu the CLI asks with — AskUserQuestion (one question or several, multiSelect too), a permission or trust prompt, Codex's approval, a first-start login menu — is answered from the chat: it is read off the terminal underneath every 700 ms (`parsePaneMenu`/`readRunMenu` in `client/src/lib/panePaste.ts`: two numbered rows or more from 1, the `❯`/`›` cursor on one of them, no `─` rule right above the first — a reply's numbered list has no cursor, a list typed into Claude's input box sits under its rule) and shown in the prompt box's place as a card of buttons (`PaneMenuCard` in `PaneChat.tsx`, with each option's description and a multiSelect box's state; the box is hidden with CSS, never unmounted, so its draft survives the question); a pick presses its digit in the pane (which picks it, or ticks its box), Claude's unnumbered **Submit** row is reached with the arrows and Enter, **Cancel (Esc)** presses Esc, and the **Type something** row opens a field inside the card whose text, on Enter, is typed there as the answer (`answerRunMenu`: the arrows to that row, Ctrl+U over any text already in it, the text as a bracketed paste, Enter — none in a multiSelect, where Submit is still to pick). That row presses no key: a second press of its digit would type the digit into the CLI's own field. Once text is typed in that row its label is the text, so it is found as the last numbered row above **Chat about this**. Like Send, it only presses keys in the pane, so no new route exists for it. A menu with no numbers (the first-start theme picker) still needs the terminal. A live pane's chat ends, until its turn is over, with what the model is doing now and — like the terminal's status line — how long the turn has run since your prompt and the output tokens it has made so far, e.g. `Thinking… (2m 8s · ↑ 3.5K tokens)` (`workingOn` and `turnStats` in `MessageList.tsx`; Claude Code repeats a call's usage on each of its lines, so consecutive lines with the same prompt size count once), the words lit by a moving shimmer (`.text-shimmer`, stopped under reduced motion) and the token count rolling up rather than jumping. It names every call of the turn still waiting for its result — the oldest first and **+N more** for the rest, a click listing them all, since Claude Code writes parallel calls on separate lines and their results arrive in any order — by its description (Claude Code's Bash `description`) or tool and target, with that call's own time once it has run 30 s; else **Thinking…**. A line with no new transcript line for 3 minutes turns amber (the step may be stuck). While the CLI asks you something — the main agent awaiting input on a `Notification` hook (a permission prompt; `awaiting_reason = 'notification'`, read with the Stop button's agent poll) — it reads **Waiting for your approval: <step>** in amber with an **Open terminal** button (`showTerminal` in `PaneGrid.tsx`, the chat/terminal toggle's own switch: the unsent prompt goes into the CLI's input box); the menu card above answers it from the chat. AskUserQuestion fires no `Notification`, so a menu read off the pane counts as waiting too — for this line and the hidden tab's **● Waiting**. Before the session has started — no session row yet, because Claude Code runs no hook until its folder is trusted — the chat says so instead of the empty-transcript text, with the same **Open terminal** button (`notStarted` in `ConversationView.tsx`), and Send and every control typed into the pane (`/model`, `/effort`) ask first (`confirmDialog`, **Send anyway**): an Enter there would answer the trust prompt, whose preselected choice is **No, exit** (issue #73). It asks rather than refuses because the row appears only with the session's first hook, and a hard lock would strand the chat if none came right after the answer. A turn that ended (`end_turn`) closes with **✓ Done in 2m 8s · ↑ 3.5K tokens** — no cost: pricing one turn would need a second cost formula. The prompt box shows how full the context is (`ContextMeter` in `PaneChat.tsx`): the newest reply's prompt size (input + cache) against the model's window — 200K for Claude's 4.5 line and older, 1M for every model since (`contextWindow`, from LiteLLM's list; none for another vendor) — amber from 75 %, red from 90 %, the numbers on hover. While the browser tab is hidden its title is marked **● Waiting** while a chat pane's session waits on you and **✓ Done** once a turn ended (`client/src/lib/tabBadge.ts`); showing the tab drops the Done mark. Thinking is shown (after your prompt, after a tool result, or after narration whose `stop_reason` is `tool_use`). The transcript route passes each assistant line's `stop_reason`; `end_turn` ends the turn, so does an interrupt, and a turn with no new line for 15 minutes and no call running is treated as dead. The chat re-reads the transcript every 3 s while the page is visible. Each message's hover line shows ↓ the whole prompt it was answered from (input + cache read + cache write; the split on hover) and ↑ its output. An AskUserQuestion is not folded into the collapsed work line: it is its own card (`AskCard`) showing each question with the option the user picked or what they typed — the `answers` the transcript route reads off Claude Code's `toolUseResult` —, "declined" when they refused it. Each assistant turn that changed files ends, like Claude Desktop's, with an **Edited N files** card (`EditedFiles` in `client/src/components/conversation/MessageList.tsx`): one row per file with the lines it added and removed over the turn — the `diff` the transcript route reads off Claude Code's `structuredPatch` — and, in a live pane's chat, a click on a row opens that file in the workspace's Changes panel when it lies in the pane's folder, else its preview (beside the chat, or over it when the pane is narrow). It has no Undo: reverting a file is a write the dashboard does not take from a click. The conversation column and the prompt box are `max(48rem, 70%)` of the pane wide — never narrower than Claude's fixed 48rem, wider on a large screen. Every chat pane, split or not, has Claude Desktop's activity panel (`client/src/components/run/ChatOutline.tsx`, `GET /api/sessions/:id/outline`, re-read every 5 s while the tab is visible). It starts open in a pane that spans the grid's full width or height (`spanning` in `ConsoleArea`, `client/src/pages/Workspace.tsx`, from `layoutRects`) once the session has an output, and otherwise — no output yet, or a quarter-sized pane (e.g. four panes at once) — collapsed into a file button with the number of outputs at the chat's top right. The user's own open or close is kept per session for the browser tab (`sessionStorage` `agenthub.chat.panel.<sessionId>`), so switching the pane to its terminal and back keeps it; a split or a close that changes the pane's shape drops it and brings the default back. Whether it sits beside the conversation or floats over it is decided by the PANE's width, not the window's (a CSS container query on `PaneChat`'s root, 40rem): beside it, a splitter on its left edge (`Divider` from `PaneGrid.tsx`, arrow keys too) drags the panel — and a file preview in its place — between 20% and 70% of the pane, remembered per browser (`localStorage` `agenthub.chat.panelWidth`, one fraction for every pane). Open, it shows **Progress** — the session's `TaskCreate`/`TaskUpdate` (or `TodoWrite`) plan, only when there is one: how many steps are done of how many (`3/7`) over a bar, done items struck through and the step in progress highlighted; **Outputs** — files it wrote or edited (`Write`/`Edit`/`MultiEdit`/`NotebookEdit`), newest first, each with its extension tag (a file made by a Bash command is not listed); and **Used in this session** — **Uploads** (files attached through the box), **Files read**, **Memory** (Claude Code memory files), **Skills** and **Connectors** (MCP servers). A click on an output opens its preview in the panel's place (`FilePreview` in `client/src/components/run/ChatFilePreview.tsx`: name · EXT, Copy path, Close back to the panel); Uploads, Files read and Memory open their list with a preview beside it (`FileListModal`), and Skills opens the same dialog over the session's skills, each previewed from its SKILL.md (`SkillListModal`: a plugin's `plugin:name` from `GET /api/cc-config/plugins`, else the user's or the project's skill; a built-in skill has no file and shows by name); Connectors opens the Claude config page. Every row but Connectors also drags onto the prompt box (`dragAttachments`/`dragSkill` in `client/src/lib/promptAttach.ts`): an output, or a whole Uploads/Files read/Memory group, attaches its files like the `+` menu does (an upload by its bare path, anything else as `@<absolute path>`); a skill becomes the message's leading `/<name>` — the Skills row only when the session used one skill (a prompt takes one command), else drag it from its list. A row dragged out of one of those list dialogs closes the dialog, so the box is reachable. A preview reads only through routes that already guard it — a file under the pane's folder through `GET /api/lanes/:id/git/file` (an image through `/git/raw`), a memory file under `~/.claude` through `GET /api/cc-config/file`, an attached image or text file (`.txt`, `.md`, `.csv`, `.log`, `.json`… — a long paste is one) through `GET /api/run/drop/:name` — and shows just the path for anything else (an attached spreadsheet, a file outside the workspace, one since deleted). A Windows path (`C:\…\.claude\…`) is matched like a POSIX one. The ended-session replay and the Analyst page's conversation use the same layout. The chat's unsent prompt and its attachments are kept per session in this browser (`agenthub.chat.draft.<sessionId>`, `client/src/lib/chatDraft.ts`), so switching the pane to its terminal and back, or reloading the page, leaves them as typed — an attached image keeps its path, not its preview; a send clears it. Switching carries the unsent prompt across: to the terminal, the draft (its files as paths) is pasted into the CLI's input box without Enter and the draft is cleared; to the chat, the text in the CLI's input box is read off the screen (`parsePromptInput` in `client/src/lib/panePaste.ts`: Claude's `❯` box between its `─` rules, Codex's last `›` row up to its footer; a soft-wrapped row is joined back) into the draft and deleted there with Delete/Backspace — never Ctrl+C or Esc, which would stop a working turn. A box holding `[Pasted text #N]` or `[Image #N]` is left as it is, since the screen does not show what it stands for. A row of a turn's **Edited N files** card drags into the prompt box as an `@path` attachment (its header drags every file in the card); dropped elsewhere it is the plain path.
+- **Chat view of a live pane** — each Claude or Codex SESSION shows either a chat or its terminal, its own choice: the chat/terminal button left of the pane's split buttons (`ChatToggle` in `client/src/components/run/PaneGrid.tsx`) switches it, and the choice is kept in this browser's localStorage by session id (`agenthub.chatUi.sessions`, the newest 200; `useSessionChat` in `client/src/lib/chatUi.ts`). **Settings → Appearance → Chat UI** (`ChatUiSettings.tsx`, saved per ACCOUNT on the server: `GET/PUT /api/me/prefs`, key `agenthub.chatUi`) is only the default — written onto a session the first time it is shown, so changing it later leaves every session already seen as it is. The chat needs the session id, so a pane shows it only once its session is known (`client/src/components/run/LaneConsolePane.tsx`). Chat (`client/src/components/run/PaneChat.tsx`) shows the session's conversation the way Claude Desktop does — the shared `ConversationView` in its `fill` layout: one centred column, your messages as bubbles, the reply as plain prose, every run of tool calls folded into one line named by what it did, in Claude Desktop's fixed order ("Ran 3 commands, edited a file, used a tool ›", `workLabel`) — a thinking block that carries text is the model's narration (Claude Code ≥ 2.1.280 stores real thinking empty) and reads as prose between those lines, as Claude Desktop shows it — that opens onto a framed list of its steps — each titled by its own description, the call in a capped scroll box, its output one click further — and a round ↓ back to the end — above a prompt box that grows with its text up to 40% of the screen, then scrolls. The conversation stays at its end while you are there: a new line, the running step, an image loading or the pane shown again after being hidden keeps the newest in sight (a `ResizeObserver` in `ConversationView`), until you scroll up; sending a message takes it back to the end. It lies OVER the terminal, which stays mounted and attached (the console never unmounts a terminal); a message is sent through that terminal as a bracketed paste and Enter (`submitToRun` in `client/src/lib/panePaste.ts`), exactly as if typed, so no new route or server write exists for it. On a Windows dashboard (psmux) the attach (`server/lib/pty-attach.js`) holds any input that follows a paste until 700 ms after it (`PASTE_HOLD_MS`): psmux delays a paste ~350 ms and wrote an Enter sent meanwhile INSIDE it (psmux#766), so the prompt sat in Claude's composer as `[Pasted text #1]` instead of being sent. While the session's main agent is `working` (its `UserPromptSubmit` hook seen, no `Stop` yet — `GET /api/agents?session_id=&status=working`, re-read on `agent_updated`), an empty box shows **Stop** in Send's place, like Claude Desktop: it presses Esc in the pane (`interruptRun` in the same file), the key `claude` and `codex` stop a turn with; typing a message brings Send back (Claude Code queues it). Stop also ends the conversation's live line (`stopped` on `ConversationView`/`MessageList`) until the transcript gains a newer line: an Esc before the model wrote anything puts the prompt back in Claude Code's input box and writes no `[Request interrupted by user]` line, so the transcript still ends on the prompt and would read as **Thinking…** for 15 minutes (issue #74). An Esc pressed in the terminal itself is not seen by the chat. A Codex pane without AgentHub's hooks never shows it. An image attached in the box reaches the transcript as its saved path; the user's turn shows the image itself (`GET /api/run/drop/:name`) — any other attached file as a chip (name · EXT) that opens its preview like an output — and opens it full size over the whole page (`ImageLightbox`, rendered into `document.body`, so nothing stacked over a pane covers it). The box carries Claude Desktop's controls: a `+` menu (`ChatAddMenu.tsx`, also in the workspace home's prompt box) — Upload or add files; what already exists, attached like an upload (a chip, its path sent with the prompt): **Project files** (a search of the box's folder, folders too — `GET /api/run/files?dirs=1` — sent as `@path`), **Context files** (the workspace's own, `GET /api/lanes/:id/sources`, sent as `@<stored path>`) and **Recent uploads** (files uploaded through a prompt box in the last day, kept in this browser's localStorage `agenthub.prompt.recentUploads` — the server prunes an upload after a day); a file or folder dragged onto anywhere in a prompt box (not only its text area) attaches the same way — from the Changes panel's Files tree, its search, a Changes-list row, or a workspace's Context files and Folder list; a row from an extra folder's section (`data-drag-root`) and a context file or folder carry their absolute path, since they lie outside the pane's folder (`client/src/lib/promptAttach.ts`); Skills (yours and the project's), Plugins → a plugin → the skills and commands it ships, each put at the front of the message as `/<name>` / `/<plugin>:<name>`, and Manage links to the Claude config page — and, for a Claude pane, the Model/Effort chip, a change typed into the pane as `/model <id>` / `/effort <level>`, and the permission mode beside `+` in the box. A menu opened in a split pane (a `Select`, the Model chip) flips up or down by the room inside ITS pane (`data-menu-bounds` on the pane, `menuRoom` in `client/src/components/Select.tsx`), never the window's — a menu running past the pane's edge is painted over by the pane below. A running session changes mode only by Shift+Tab, so the server steps it and reads each step back off the pane (`POST /api/run/:id/permission-mode`, `server/lib/pane-mode.js`): it lands on exactly the mode picked, presses nothing while a menu hides the mode line, and returns to where it began when the mode is not in that session's cycle (Bypass only when the session was started with it). The mode belongs to that session only. A menu the CLI asks with — AskUserQuestion (one question or several, multiSelect too), a permission or trust prompt, Codex's approval, a first-start login menu — is answered from the chat: it is read off the terminal underneath every 700 ms while the pane is on screen (`parsePaneMenu`/`readRunMenu` in `client/src/lib/panePaste.ts`: two numbered rows or more from 1, the `❯`/`›` cursor on one of them, no `─` rule right above the first — a reply's numbered list has no cursor, a list typed into Claude's input box sits under its rule) and shown in the prompt box's place as a card of buttons (`PaneMenuCard` in `PaneChat.tsx`, with each option's description and a multiSelect box's state; the box is hidden with CSS, never unmounted, so its draft survives the question); a pick presses its digit in the pane (which picks it, or ticks its box), Claude's unnumbered **Submit** row is reached with the arrows and Enter, **Cancel (Esc)** presses Esc, and the **Type something** row opens a field inside the card whose text, on Enter, is typed there as the answer (`answerRunMenu`: the arrows to that row, Ctrl+U over any text already in it, the text as a bracketed paste, Enter — none in a multiSelect, where Submit is still to pick). That row presses no key: a second press of its digit would type the digit into the CLI's own field. Once text is typed in that row its label is the text, so it is found as the last numbered row above **Chat about this**. Like Send, it only presses keys in the pane, so no new route exists for it. A menu with no numbers (the first-start theme picker) still needs the terminal. A live pane's chat ends, until its turn is over, with what the model is doing now and — like the terminal's status line — how long the turn has run since your prompt and the output tokens it has made so far, e.g. `Thinking… (2m 8s · ↑ 3.5K tokens)` (`workingOn` and `turnStats` in `MessageList.tsx`; Claude Code repeats a call's usage on each of its lines, so consecutive lines with the same prompt size count once), the words lit by a shimmer that sweeps a few times and rests (`.text-shimmer`, stopped under reduced motion) and the token count rolling up rather than jumping. It names every call of the turn still waiting for its result — the oldest first and **+N more** for the rest, a click listing them all, since Claude Code writes parallel calls on separate lines and their results arrive in any order — by its description (Claude Code's Bash `description`) or tool and target, with that call's own time once it has run 30 s; else **Thinking…**. A line with no new transcript line for 3 minutes turns amber (the step may be stuck). While the CLI asks you something — the main agent awaiting input on a `Notification` hook (a permission prompt; `awaiting_reason = 'notification'`, read with the Stop button's agent poll) — it reads **Waiting for your approval: <step>** in amber with an **Open terminal** button (`showTerminal` in `PaneGrid.tsx`, the chat/terminal toggle's own switch: the unsent prompt goes into the CLI's input box); the menu card above answers it from the chat. AskUserQuestion fires no `Notification`, so a menu read off the pane counts as waiting too — for this line and the hidden tab's **● Waiting**. Before the session has started — no session row yet, because Claude Code runs no hook until its folder is trusted — the chat says so instead of the empty-transcript text, with the same **Open terminal** button (`notStarted` in `ConversationView.tsx`), and Send and every control typed into the pane (`/model`, `/effort`) ask first (`confirmDialog`, **Send anyway**): an Enter there would answer the trust prompt, whose preselected choice is **No, exit** (issue #73). It asks rather than refuses because the row appears only with the session's first hook, and a hard lock would strand the chat if none came right after the answer. A turn that ended (`end_turn`) closes with **✓ Done in 2m 8s · ↑ 3.5K tokens** — no cost: pricing one turn would need a second cost formula. The prompt box shows how full the context is (`ContextMeter` in `PaneChat.tsx`): the newest reply's prompt size (input + cache) against the model's window — 200K for Claude's 4.5 line and older, 1M for every model since (`contextWindow`, from LiteLLM's list; none for another vendor) — amber from 75 %, red from 90 %, the numbers on hover. While the browser tab is hidden its title is marked **● Waiting** while a chat pane's session waits on you and **✓ Done** once a turn ended (`client/src/lib/tabBadge.ts`); showing the tab drops the Done mark. Thinking is shown (after your prompt, after a tool result, or after narration whose `stop_reason` is `tool_use`). The transcript route passes each assistant line's `stop_reason`; `end_turn` ends the turn, so does an interrupt, and a turn with no new line for 15 minutes and no call running is treated as dead. The chat re-reads the transcript every 3 s while the page is visible and the pane is on screen (`useOnScreen` in `client/src/hooks/useOnScreen.ts`: a pane of another workspace or tab stays mounted but `display:none`, and polls nothing until it is shown again, then catches up once), and on a `new_event` of its session at most every 1.5 s (`EVENT_FETCH_MS`). The tail read (`?after=`) skips the lines the chat already has before parsing them. While the chat covers the terminal, the terminal is moved out of the viewport with a transform (`covered` on `TerminalView`), which stops xterm drawing without changing its size — the menu reading below still works. Each message's hover line shows ↓ the whole prompt it was answered from (input + cache read + cache write; the split on hover) and ↑ its output. An AskUserQuestion is not folded into the collapsed work line: it is its own card (`AskCard`) showing each question with the option the user picked or what they typed — the `answers` the transcript route reads off Claude Code's `toolUseResult` —, "declined" when they refused it. Each assistant turn that changed files ends, like Claude Desktop's, with an **Edited N files** card (`EditedFiles` in `client/src/components/conversation/MessageList.tsx`): one row per file with the lines it added and removed over the turn — the `diff` the transcript route reads off Claude Code's `structuredPatch` — and, in a live pane's chat, a click on a row opens that file in the workspace's Changes panel when it lies in the pane's folder, else its preview (beside the chat, or over it when the pane is narrow). It has no Undo: reverting a file is a write the dashboard does not take from a click. The conversation column and the prompt box are `max(48rem, 70%)` of the pane wide — never narrower than Claude's fixed 48rem, wider on a large screen. Every chat pane, split or not, has Claude Desktop's activity panel (`client/src/components/run/ChatOutline.tsx`, `GET /api/sessions/:id/outline`, re-read every 5 s while the tab is visible and the pane is on screen; the server keeps the last answer per session until its events change). It starts open in a pane that spans the grid's full width or height (`spanning` in `ConsoleArea`, `client/src/pages/Workspace.tsx`, from `layoutRects`) once the session has an output, and otherwise — no output yet, or a quarter-sized pane (e.g. four panes at once) — collapsed into a file button with the number of outputs at the chat's top right. The user's own open or close is kept per session for the browser tab (`sessionStorage` `agenthub.chat.panel.<sessionId>`), so switching the pane to its terminal and back keeps it; a split or a close that changes the pane's shape drops it and brings the default back. Whether it sits beside the conversation or floats over it is decided by the PANE's width, not the window's (a CSS container query on `PaneChat`'s root, 40rem): beside it, a splitter on its left edge (`Divider` from `PaneGrid.tsx`, arrow keys too) drags the panel — and a file preview in its place — between 20% and 70% of the pane, remembered per browser (`localStorage` `agenthub.chat.panelWidth`, one fraction for every pane). Open, it shows **Progress** — the session's `TaskCreate`/`TaskUpdate` (or `TodoWrite`) plan, only when there is one: how many steps are done of how many (`3/7`) over a bar, done items struck through and the step in progress highlighted; **Outputs** — files it wrote or edited (`Write`/`Edit`/`MultiEdit`/`NotebookEdit`), newest first, each with its extension tag (a file made by a Bash command is not listed); and **Used in this session** — **Uploads** (files attached through the box), **Files read**, **Memory** (Claude Code memory files), **Skills** and **Connectors** (MCP servers). A click on an output opens its preview in the panel's place (`FilePreview` in `client/src/components/run/ChatFilePreview.tsx`: name · EXT, Copy path, Close back to the panel); Uploads, Files read and Memory open their list with a preview beside it (`FileListModal`), and Skills opens the same dialog over the session's skills, each previewed from its SKILL.md (`SkillListModal`: a plugin's `plugin:name` from `GET /api/cc-config/plugins`, else the user's or the project's skill; a built-in skill has no file and shows by name); Connectors opens the Claude config page. Every row but Connectors also drags onto the prompt box (`dragAttachments`/`dragSkill` in `client/src/lib/promptAttach.ts`): an output, or a whole Uploads/Files read/Memory group, attaches its files like the `+` menu does (an upload by its bare path, anything else as `@<absolute path>`); a skill becomes the message's leading `/<name>` — the Skills row only when the session used one skill (a prompt takes one command), else drag it from its list. A row dragged out of one of those list dialogs closes the dialog, so the box is reachable. A preview reads only through routes that already guard it — a file under the pane's folder through `GET /api/lanes/:id/git/file` (an image through `/git/raw`), a memory file under `~/.claude` through `GET /api/cc-config/file`, an attached image or text file (`.txt`, `.md`, `.csv`, `.log`, `.json`… — a long paste is one) through `GET /api/run/drop/:name` — and shows just the path for anything else (an attached spreadsheet, a file outside the workspace, one since deleted). A Windows path (`C:\…\.claude\…`) is matched like a POSIX one. The ended-session replay and the Analyst page's conversation use the same layout. The chat's unsent prompt and its attachments are kept per session in this browser (`agenthub.chat.draft.<sessionId>`, `client/src/lib/chatDraft.ts`), so switching the pane to its terminal and back, or reloading the page, leaves them as typed — an attached image keeps its path, not its preview; a send clears it. Switching carries the unsent prompt across: to the terminal, the draft (its files as paths) is pasted into the CLI's input box without Enter and the draft is cleared; to the chat, the text in the CLI's input box is read off the screen (`parsePromptInput` in `client/src/lib/panePaste.ts`: Claude's `❯` box between its `─` rules, Codex's last `›` row up to its footer; a soft-wrapped row is joined back) into the draft and deleted there with Delete/Backspace — never Ctrl+C or Esc, which would stop a working turn. A box holding `[Pasted text #N]` or `[Image #N]` is left as it is, since the screen does not show what it stands for. A row of a turn's **Edited N files** card drags into the prompt box as an `@path` attachment (its header drags every file in the card); dropped elsewhere it is the plain path.
 - **The workspace home** — `/workspace?lane=<id>` opens on the workspace's home (`client/src/components/lanes/WorkspaceHome.tsx`), shaped like a Claude Desktop project page: the name (rename from its `⋯` menu, which is the sidebar row menu) with the repository it belongs to under it (root, the main worktree's branch, the worktree count, from `GET /api/projects` — the facts Settings → Projects used to show), a prompt box — the one place a session starts WITH a prompt (a pane has no run form; the session view's tab bar `+` only starts a bare CLI or a shell) — whose fields run in order **AI** (the session's own tool: every installed, enabled one, the workspace's by default), **AI Proxy** (Native or a saved profile; Native alone, disabled, for a tool that cannot use one), **Model** (the tool's own list, or on a router that router's models), **Permission** (Claude's permission mode; Codex's sandbox and approval; a catalog tool's per-session bypass switch) and **Effort** (Claude and Codex), whose half-typed text is kept per workspace in `agenthub.workspace.draft.<lane id>` (localStorage, so it survives leaving the page or switching workspace; cleared once sent — attached images are not kept), and whose **Start** opens a NEW session in the first slot with no live run (0–9, `POST /api/lanes/:id/start {slot}`; all ten busy is refused) and then shows it in a tab of its own, and **Recents** — the workspace's own sessions (`?lane=<id>`) with their cost and time, a live one opening its pane and any other its conversation (`&session=`), exactly like the sidebar. The transcript page (`/sessions/<id>`) of a Claude session started in a workspace's own folder carries an explicit **Resume** (`client/src/components/lanes/SessionResumeButton.tsx`, the one place that resumes) and beside it **Fork**, which resumes the history into a NEW session id (`{forkSession: true}` → `--fork-session`, the original untouched, offered even while the original has a live pane): it warns that a resume re-reads the whole transcript into context, re-reads the live panes, then starts it in the workspace's first free pane (`POST /api/lanes/:id/start {slot, resumeSessionId}`) and opens it. A session from a SUBFOLDER gets no Resume — the pane starts in the workspace's folder and Claude finds a session by the folder it ran in — and a session that still has a live pane shows **Open pane** instead. On the right sit the workspace's own context: **Instructions** (appended to every new session's system prompt, after the workspace owner's own **My agent instructions** from Settings → AI tools — one markdown text per account (typed, or read from a local `.md`/`.txt` with **Choose file**, saved only on Save), shared by every workspace they own), **Memory** (read-only: the folder's CLAUDE.md / AGENTS.md family and Claude's auto memory), **Context** (uploaded files, ≤ 10 MB each, stored by AgentHub) and **Folder** (the workspace's own folder plus extra ones, with the repo's untracked worktrees suggested). Only Claude and Codex panes can take them (`--add-dir`, `--append-system-prompt-file=` / `-c developer_instructions=`); for any other tool the page says so and its panes start without them. See `docs/API.md` → Workspace context. Each extra folder is also its own repo in the **Changes** panel, stacked under the workspace folder's like VS Code's Source Control: once the workspace has one, every repo is a section with a foldable header, each a whole panel — Changes, Files, Review, History, commit and push — on its own folder (every git route with `?source=<id>`, resolved from the `lane_sources` row; `ENOSOURCE` otherwise, never the workspace folder); the open sections share the height, and a folded one stays mounted. A chat's request to open a file unfolds the workspace folder's section. Right-clicking a section header offers **Close repository** (extra folders only — the workspace folder never closes) and **Close other repositories**; closing REMOVES the folder from the workspace (`DELETE /api/lanes/:id/sources/:sourceId`, the folder stays on disk), asking first only if that section holds an unsaved edit. The panel's bottom fold, **Workspace context** (folded by default, kept in `agenthub.workspace.contextOpen`), is the home's context column made compact (`client/src/components/lanes/WorkspaceContext.tsx`, the same component): Instructions, Memory, Context files and Folders are edited from the session view without going back to the home, and a folder added or removed there joins or leaves the stack above at once. A plain-text paste over 5,000 characters into any prompt box (the home, a chat, a finished session's reply) is attached as `pasted-text.txt` like a pasted file (`PASTE_AS_FILE_CHARS` in `client/src/components/run/RunSetup.tsx`): the prompt carries its path, never the text (#68). The home HIDES the session view with CSS — the console never unmounts behind it.
 - **A session in its own worktree** — Claude Desktop's worktree toggle: the row under the home's prompt box has a **Local / New worktree** select (Claude workspaces in a git repo). Claude Code itself makes a worktree only when started with `--worktree` or when asked mid-session — and a worktree it enters mid-session is invisible to AgentHub, which reads a session's folder once, at its start, and the new session's pane runs `claude --worktree=<name>`, so CLAUDE CODE makes `<repo>/.claude/worktrees/<name>` on branch `worktree-<name>`, copies `.worktreeinclude`, runs your `WorktreeCreate` hook, and on exit removes a worktree with no changes (or asks). AgentHub adds `/.claude/worktrees/` to the repo's `info/exclude` (Claude does not, and the main repo would list it). The session stays this workspace's own — listed with a branch icon and the worktree's name — and the session view's Changes panel works in ITS worktree (every git route with `?session=`, resolved from the session row: owned by the workspace, registered by `git worktree list`; `ENOWORKTREE` otherwise, never the main folder). Resume and Fork start in that worktree. **Archive** of a worktree session with no live pane asks first and removes the worktree like Claude Desktop — `git worktree remove` without force (git refuses unsaved work, and then nothing is archived) and `git branch -d` (unmerged work keeps its branch); a lock held by a live process refuses, a stale one is cleared. The workspace's own folder is never removed.
 - **Resume interrupted sessions** — when a session's pane dies under it (the machine shut down, tmux was killed, the CLI crashed), its run is recorded `interrupted` instead of `abandoned`, and every page of the dashboard opens the **Resume interrupted sessions?** dialog (`client/src/components/lanes/InterruptedSessionsPrompt.tsx`, from `GET /api/run/interrupted`). Each session is listed with its workspace, tool, name or first prompt and when it stopped, all ticked: **Resume** starts each ticked one in its workspace's first free tab (`POST /api/lanes/:id/start {slot: "auto", resumeSessionId}`, with the run's model, effort and permission mode — never `bypassPermissions`) and opens the first; **Don't resume** (also ×, Escape, the backdrop) resumes none. Either answer dismisses the whole list (`POST /api/run/interrupted/dismiss`), so it is asked once; a resume that failed stays listed. What counts as interrupted: a pane the dashboard did not close (Stop, ×, Quit are `killed`), not a Terminal tab, with a known session — and for Claude, not one the user ended themselves (`SessionEnd` reason `prompt_input_exit`/`logout`). Codex and other tools report no such reason, so their every vanished pane counts; a Codex pane without its opt-in hooks has no known session and is never listed. A session resumed by any other path drops off the list. A run younger than 20 s is never judged dead (a tmux list is a cached answer, and another dashboard on the same database may not have listed a brand-new pane yet). In server mode each user is offered only their own workspaces' sessions.
diff --git a/docs/MONITORING.md b/docs/MONITORING.md
index 81e96713..612d2cf3 100644
--- a/docs/MONITORING.md
+++ b/docs/MONITORING.md
@@ -24,7 +24,8 @@ as they are into an existing Grafana (Dashboards → New → Import).
 ## Metrics used
 
 `agenthub_up`, `agenthub_build_info{version}`, `agenthub_process_uptime_seconds`,
-`agenthub_process_resident_memory_bytes`, `agenthub_sessions{status}`,
+`agenthub_process_resident_memory_bytes`, `agenthub_event_loop_delay_seconds{quantile}`,
+`agenthub_sessions{status}`,
 `agenthub_agents{status}`, `agenthub_events_total`, `agenthub_websocket_clients`,
 `agenthub_remote_sources{enabled}`, `agenthub_tokens_total{kind}`.
 
diff --git a/docs/SERVER-MODE.md b/docs/SERVER-MODE.md
index 339ca617..bfe3e340 100644
--- a/docs/SERVER-MODE.md
+++ b/docs/SERVER-MODE.md
@@ -796,7 +796,9 @@ What a new machine still needs by hand after it:
    `PostToolUse` (both with `"matcher": "*"`), `Stop`, `SubagentStop`,
    `Notification`, `SessionStart`, `SessionEnd` and `UserPromptSubmit`, a
    `command` hook `/opt/agenthub/libexec/member-hook <Event>` with
-   `"timeout": 5`.
+   `"timeout": 5` — and `"async": true` on `PreToolUse` and `PostToolUse`, so
+   Claude Code does not wait a Node start before and after every tool call.
+   Re-running `install-system` rewrites a managed file it wrote itself.
 5. Any other daemon that must see every process (monitoring agents, for
    example) needs group `agenthub-proc`: `sudo usermod -aG agenthub-proc <it>`.
 6. A tool `agenthub server tools` reports as `ESHAREDLAYOUT` (a script in a
diff --git a/scripts/install-hooks.js b/scripts/install-hooks.js
index 93e0bde4..0d447438 100644
--- a/scripts/install-hooks.js
+++ b/scripts/install-hooks.js
@@ -127,6 +127,12 @@ const HOOKS_WITH_MATCHER = ["PreToolUse", "PostToolUse", "Stop", "SubagentStop",
 // tool-name matchers, hence the separate list.
 const HOOKS_WITHOUT_MATCHER = ["SessionStart", "SessionEnd", "UserPromptSubmit"];
 const HOOK_TYPES = [...HOOKS_WITH_MATCHER, ...HOOKS_WITHOUT_MATCHER];
+// Fired around EVERY tool call, and Claude Code waits for a synchronous hook
+// before it goes on — each one a Node start (0.1–0.2 s), so ~0.4 s added to
+// every tool call. The handler only reports (it prints nothing, decides
+// nothing), so these run in the background (`async`). The rarer events stay
+// synchronous so a turn's start and end still reach the dashboard in order.
+const ASYNC_HOOKS = new Set(["PreToolUse", "PostToolUse"]);
 
 function makeHookEntry(hookType) {
   const entry = {
@@ -134,6 +140,7 @@ function makeHookEntry(hookType) {
       {
         type: "command",
         command: `node "${HOOK_HANDLER}" ${hookType}`,
+        ...(ASYNC_HOOKS.has(hookType) && { async: true }),
       },
     ],
   };
@@ -247,4 +254,12 @@ if (require.main === module) {
 
 // isOurEntry is also used by scripts/plugin-bootstrap.js to strip hook entries
 // a previous checkout install left behind (they would double-count events).
-module.exports = { HOOK_TYPES, installHooks, isInsideContainer, isOurEntry, pluginProvidesHooks };
+module.exports = {
+  ASYNC_HOOKS,
+  HOOK_TYPES,
+  installHooks,
+  isInsideContainer,
+  isOurEntry,
+  makeHookEntry,
+  pluginProvidesHooks,
+};
diff --git a/scripts/server/install-system.sh b/scripts/server/install-system.sh
index b66f0439..3c2a5495 100644
--- a/scripts/server/install-system.sh
+++ b/scripts/server/install-system.sh
@@ -156,8 +156,11 @@ if [ -f "$MANAGED" ] && ! { [ -f "$MANAGED_MARK" ] && sha256sum -c --status "$MA
 else
   "$NODE" -e '
     const ev = ["PreToolUse","PostToolUse","Stop","SubagentStop","Notification","SessionStart","SessionEnd","UserPromptSubmit"];
+    // Per-tool-call hooks run in the background (async): a synchronous one
+    // makes Claude Code wait a Node start (0.1-0.2 s) before and after every call.
     const hook = (e) => [{ ...(["PreToolUse","PostToolUse"].includes(e) ? { matcher: "*" } : {}),
-      hooks: [{ type: "command", command: `/opt/agenthub/libexec/member-hook ${e}`, timeout: 5 }] }];
+      hooks: [{ type: "command", command: `/opt/agenthub/libexec/member-hook ${e}`, timeout: 5,
+        ...(["PreToolUse","PostToolUse"].includes(e) ? { async: true } : {}) }] }];
     process.stdout.write(JSON.stringify({ hooks: Object.fromEntries(ev.map((e) => [e, hook(e)])) }, null, 2) + "\n");
   ' >"$MANAGED"
   chmod 0644 "$MANAGED"
diff --git a/server/__tests__/daily-cost.test.js b/server/__tests__/daily-cost.test.js
index 28211846..de47f59c 100644
--- a/server/__tests__/daily-cost.test.js
+++ b/server/__tests__/daily-cost.test.js
@@ -234,6 +234,46 @@ describe("TranscriptCache tokensByHour", () => {
   });
 });
 
+describe("writeHourly skips unchanged hours", () => {
+  const { writeHourly, forgetHourly } = require("../lib/hourly-usage");
+  const bucket = (input) => ({
+    hour: "2026-10-05T08",
+    model: "claude-opus-4-1",
+    speed: "standard",
+    geo: "",
+    tier: "standard",
+    input,
+    output: 0,
+    cacheRead: 0,
+    cacheWrite: 0,
+    cacheWrite1h: 0,
+    webSearch: 0,
+    webFetch: 0,
+    codeExec: 0,
+  });
+  const rows = () =>
+    db
+      .prepare("SELECT input_tokens FROM token_usage_hourly WHERE session_id = 'skip-hourly'")
+      .all()
+      .map((r) => r.input_tokens);
+
+  it("writes a changed hour, skips an unchanged one, and writes again once forgotten", () => {
+    session("skip-hourly", "2026-10-05T08:00:00.000Z");
+    writeHourly("skip-hourly", { a: bucket(5) });
+    assert.deepEqual(rows(), [5]);
+    writeHourly("skip-hourly", { a: bucket(9) });
+    assert.deepEqual(rows(), [9]);
+    // Deleted behind the writer's back: an unchanged bucket is not re-sent…
+    db.prepare("DELETE FROM token_usage_hourly WHERE session_id = 'skip-hourly'").run();
+    writeHourly("skip-hourly", { a: bucket(9) });
+    assert.deepEqual(rows(), []);
+    // …which is why every path that deletes hourly rows forgets them.
+    forgetHourly(["skip-hourly"]);
+    writeHourly("skip-hourly", { a: bucket(9) });
+    assert.deepEqual(rows(), [9]);
+  });
+});
+
 describe("session list cost", () => {
   it("equals the session's own cost, speed and server-tool fees included", async () => {
     session("list-fast", "2026-10-01T00:00:00.000Z");
diff --git a/server/__tests__/install-hooks.test.js b/server/__tests__/install-hooks.test.js
index dd0168fd..b7d6f724 100644
--- a/server/__tests__/install-hooks.test.js
+++ b/server/__tests__/install-hooks.test.js
@@ -142,6 +142,19 @@ describe("install-hooks host-only guard (#193)", () => {
   });
 });
 
+describe("async hooks", () => {
+  it("runs only the per-tool-call hooks in the background, in settings and the plugin alike", () => {
+    const { makeHookEntry, ASYNC_HOOKS } = require("../../scripts/install-hooks");
+    const plugin = require("../../.claude-plugin/plugin.json");
+    assert.deepEqual([...ASYNC_HOOKS].sort(), ["PostToolUse", "PreToolUse"]);
+    for (const type of HOOK_TYPES) {
+      const want = ASYNC_HOOKS.has(type) ? true : undefined;
+      assert.equal(makeHookEntry(type).hooks[0].async, want, `settings ${type}`);
+      assert.equal(plugin.hooks[type][0].hooks[0].async, want, `plugin ${type}`);
+    }
+  });
+});
+
 describe("HOOK_TYPES", () => {
   it("is the event list the plugin ships, so the hook status checks every one", () => {
     const exported = require("../../scripts/install-hooks").HOOK_TYPES;
diff --git a/server/__tests__/metrics.test.js b/server/__tests__/metrics.test.js
index b9b7cf41..de9bac75 100644
--- a/server/__tests__/metrics.test.js
+++ b/server/__tests__/metrics.test.js
@@ -93,6 +93,7 @@ describe("GET /api/metrics", () => {
       "agenthub_build_info",
       "agenthub_process_uptime_seconds",
       "agenthub_process_resident_memory_bytes",
+      "agenthub_event_loop_delay_seconds",
       "agenthub_sessions",
       "agenthub_agents",
       "agenthub_events_total",
diff --git a/server/__tests__/tmux.test.js b/server/__tests__/tmux.test.js
index 4034f188..2ee73af1 100644
--- a/server/__tests__/tmux.test.js
+++ b/server/__tests__/tmux.test.js
@@ -19,6 +19,21 @@ describe("tmux wrapper", () => {
     assert.equal(tmux.hasSession("agenthub-lane-1"), true);
   });
 
+  it("hasSession answers yes from a fresh list, and asks tmux for anything else", () => {
+    const calls = [];
+    tmux.__setExecImpl((args) => {
+      calls.push(args[0]);
+      if (args[0] === "list-sessions") return "agenthub-lane-1\nagenthub-lane-1-2\n";
+      throw Object.assign(new Error("no session"), { status: 1 });
+    });
+    tmux.listSessions("agenthub-lane-");
+    assert.equal(tmux.hasSession("agenthub-lane-1-2"), true);
+    assert.deepEqual(calls, ["list-sessions"]); // no has-session spawn
+    // Not listed: asked for real (a pane that just died is never trusted).
+    assert.equal(tmux.hasSession("agenthub-lane-1-3"), false);
+    assert.deepEqual(calls, ["list-sessions", "has-session"]);
+  });
+
   it("hasSession returns false when execFileSync throws", () => {
     tmux.__setExecImpl(() => {
       const e = new Error("no such session");
diff --git a/server/lib/hourly-usage.js b/server/lib/hourly-usage.js
index 57977391..a68a2882 100644
--- a/server/lib/hourly-usage.js
+++ b/server/lib/hourly-usage.js
@@ -11,11 +11,25 @@ const { db, stmts } = require("../db");
 const TranscriptCache = require("./transcript-cache");
 const { longArgs } = require("./token-usage");
 
-/** Write every hour of one transcript parse (MAX per hour, see the statement). */
+// What this process last wrote per session and hour bucket: a long session has
+// one bucket per active hour per model, and every hook re-sent all of them.
+// The upsert keeps the MAX, so skipping a bucket whose numbers did not move
+// writes exactly the same rows. Bounded (oldest session dropped first); a miss
+// only means one full write, never a lost one.
+const WRITTEN_MAX = 500;
+const written = new Map();
+
+/** Write the hours of one transcript parse that changed (MAX per hour, see the statement). */
 function writeHourly(sessionId, tokensByHour) {
-  // Runs every hour of the session on each hook — ~1 row per active hour per
-  // model; the statement leaves a row that did not grow untouched (no write).
-  for (const t of Object.values(tokensByHour || {})) {
+  let seen = written.get(sessionId);
+  if (seen) written.delete(sessionId);
+  else seen = new Map();
+  written.set(sessionId, seen);
+  if (written.size > WRITTEN_MAX) written.delete(written.keys().next().value);
+  for (const [key, t] of Object.entries(tokensByHour || {})) {
+    const sig = JSON.stringify(t);
+    if (seen.get(key) === sig) continue;
+    seen.set(key, sig);
     stmts.upsertHourlyUsage.run(
       sessionId,
       t.hour,
@@ -73,4 +87,11 @@ function backfillHourly() {
   });
 }
 
-module.exports = { writeHourly, backfillHourly };
+/** Forget what was written — for `ids`, or every session. Whatever deletes
+ *  hourly rows calls it, or the next hooks would skip re-writing them. */
+function forgetHourly(ids = null) {
+  if (!ids) return written.clear();
+  for (const id of ids) written.delete(id);
+}
+
+module.exports = { writeHourly, backfillHourly, forgetHourly };
diff --git a/server/lib/pane-mode.js b/server/lib/pane-mode.js
index 1ca2c8e6..0b3df564 100644
--- a/server/lib/pane-mode.js
+++ b/server/lib/pane-mode.js
@@ -57,13 +57,18 @@ function readMode(runId) {
 const fail = (code, message) => Object.assign(new Error(message), { code });
 const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
 
+// Between two reads of the pane. Each read is a SYNCHRONOUS capture-pane
+// (through sudo for a member's pane) that holds up every other request and
+// socket while it runs, so a mode change polls gently rather than 20×/s.
+const POLL_MS = 150;
+
 // The mode once it differs from `from` — or what the pane shows at the deadline.
 async function nextMode(runId, from, waitMs) {
   const until = Date.now() + waitMs;
   for (;;) {
     const now = readMode(runId);
     if ((now && now !== from) || Date.now() >= until) return now;
-    await sleep(50);
+    await sleep(POLL_MS);
   }
 }
 
diff --git a/server/lib/tmux.js b/server/lib/tmux.js
index 6a4e7c61..bbdbd097 100644
--- a/server/lib/tmux.js
+++ b/server/lib/tmux.js
@@ -352,7 +352,21 @@ const session = (name) => (exact() ? `=${name}` : name);
 const pane = (name) => (exact() ? `=${name}:` : name);
 
 function hasSession(name) {
+  // A fresh shared list that names it answers "yes" without a spawn: each
+  // synchronous one blocks every socket (terminal echo included) — 50-150 ms
+  // per psmux call on Windows — and a pane-token request, an AI-proxy call or
+  // a lane broadcast asks this every time. "No" is always asked for real, so a
+  // pane that just died is never trusted past the list's 1 s TTL.
+  const hit = listed.get("");
   try {
+    if (
+      hit &&
+      !hit.err &&
+      Date.now() - hit.at < LIST_TTL_MS &&
+      hit.out.split("\n").includes(name) &&
+      !accountOf(name)
+    )
+      return true;
     onServerOf(name, ["has-session", "-t", session(name)]);
     return true;
   } catch {
diff --git a/server/routes/lanes.js b/server/routes/lanes.js
index 8585ab54..6cf89bce 100644
--- a/server/routes/lanes.js
+++ b/server/routes/lanes.js
@@ -209,7 +209,11 @@ function featurePayload(feature) {
 /** Push the current state of one lane to every connected client. */
 function broadcastLane(id) {
   const lane = lanesLib.getLane(id);
-  if (lane) broadcast("lane_update", { lane: payload(lane) });
+  if (!lane) return;
+  // The cached list-sessions (as GET / uses), not a synchronous has-session —
+  // through sudo for a member's pane — on every hook that moves a lane.
+  const live = lane.run_id ? tmux.liveSessions() : null;
+  broadcast("lane_update", { lane: payload(lane, live) });
 }
 
 router.get("/", (req, res) => {
diff --git a/server/routes/metrics.js b/server/routes/metrics.js
index 1c94432c..e0f8db39 100644
--- a/server/routes/metrics.js
+++ b/server/routes/metrics.js
@@ -29,6 +29,22 @@ const { calculateCost, dailyTokenRows } = require("./pricing");
 
 const router = Router();
 
+// How long the event loop was held up — what an operator sees as a laggy
+// terminal: every keystroke's echo waits for it. Sampled every 20 ms and reset
+// each minute, so the gauges describe the last minute or so.
+const LOOP_WINDOW_MS = 60_000;
+const loopDelay = require("perf_hooks").monitorEventLoopDelay({ resolution: 20 });
+loopDelay.enable();
+let loopWindow = null; // the last full minute, reported until the next one ends
+setInterval(() => {
+  loopWindow = {
+    p50: loopDelay.percentile(50),
+    p99: loopDelay.percentile(99),
+    max: loopDelay.max,
+  };
+  loopDelay.reset();
+}, LOOP_WINDOW_MS).unref();
+
 // Resolved once at load — the dashboard version, surfaced as a build-info label
 // (the canonical Prometheus idiom for exposing a version string as a metric).
 const APP_VERSION = (() => {
@@ -100,6 +116,23 @@ router.get("/", (_req, res) => {
     [{ value: process.memoryUsage().rss }]
   );
 
+  const loop = loopWindow || {
+    p50: loopDelay.percentile(50),
+    p99: loopDelay.percentile(99),
+    max: loopDelay.max,
+  };
+  appendMetric(
+    out,
+    "agenthub_event_loop_delay_seconds",
+    "Event-loop delay over the last minute (p50, p99, max): how long requests, hooks and terminal echo waited.",
+    "gauge",
+    [
+      { labels: { quantile: "0.5" }, value: loop.p50 / 1e9 },
+      { labels: { quantile: "0.99" }, value: loop.p99 / 1e9 },
+      { labels: { quantile: "1" }, value: loop.max / 1e9 },
+    ]
+  );
+
   // Sessions by status.
   const sessionCounts = new Map(stmts.sessionStatusCounts.all().map((r) => [r.status, r.count]));
   appendMetric(
diff --git a/server/routes/sessions.js b/server/routes/sessions.js
index 73a04291..113b55de 100644
--- a/server/routes/sessions.js
+++ b/server/routes/sessions.js
@@ -469,11 +469,31 @@ function sessionOutline(sessionId) {
   };
 }
 
+// Each open chat polls its outline every few seconds, and building it parses
+// every tool event of the session: keep the last answer per session until its
+// events change (a new id, or fewer rows after a cleanup).
+const outlineStamp = db.prepare(
+  "SELECT MAX(id) AS max_id, COUNT(*) AS n FROM events WHERE session_id = ?"
+);
+const OUTLINE_CACHE_MAX = 100;
+const outlineCache = new Map();
+function cachedOutline(sessionId) {
+  const st = outlineStamp.get(sessionId);
+  const stamp = `${st.max_id}:${st.n}`;
+  const hit = outlineCache.get(sessionId);
+  if (hit && hit.stamp === stamp) return hit.body;
+  const body = sessionOutline(sessionId);
+  outlineCache.delete(sessionId);
+  outlineCache.set(sessionId, { stamp, body });
+  if (outlineCache.size > OUTLINE_CACHE_MAX) outlineCache.delete(outlineCache.keys().next().value);
+  return body;
+}
+
 router.get("/:id/outline", (req, res) => {
   if (!stmts.getSession.get(req.params.id)) {
     return res.status(404).json({ error: { code: "NOT_FOUND", message: "Session not found" } });
   }
-  res.json(sessionOutline(req.params.id));
+  res.json(cachedOutline(req.params.id));
 });
 
 // An image the session's own Read tool returned (`?path=`, exact), for the Chat
@@ -1165,10 +1185,13 @@ router.get("/:id/transcript", async (req, res) => {
     }
 
     if (afterLine !== null) {
-      // Incremental mode: skip lines until after afterLine, collect up to limit, then stop
-      let foundStart = false;
+      // Incremental mode: skip lines until after afterLine, collect up to limit, then stop.
+      // A line the client already has is skipped BEFORE it is parsed: every open
+      // chat tails its transcript every few seconds, and parsing the whole JSONL
+      // (MBs for a long session) each time was most of the server's steady CPU.
       for await (const line of rl) {
         lineNum++;
+        if (lineNum <= afterLine) continue;
         if (!line.trim()) continue;
         let entry;
         try {
@@ -1178,11 +1201,6 @@ router.get("/:id/transcript", async (req, res) => {
         }
         if (!renders(entry)) continue;
 
-        if (!foundStart) {
-          if (lineNum <= afterLine) continue;
-          foundStart = true;
-        }
-
         const message = isCodex ? codexMessage(entry, lineNum) : parseMessage(entry, lineNum);
         if (!message) continue;
         total++;
diff --git a/server/routes/settings.js b/server/routes/settings.js
index 56b26630..ea67d61d 100644
--- a/server/routes/settings.js
+++ b/server/routes/settings.js
@@ -11,6 +11,7 @@ const v8 = require("v8");
 const { db, stmts, DB_PATH, DEFAULT_PRICING, applyIntroPricing } = require("../db");
 const { getConnectionCount } = require("../websocket");
 const { transcriptCache } = require("./hooks");
+const { forgetHourly } = require("../lib/hourly-usage");
 
 const router = Router();
 
@@ -179,6 +180,7 @@ router.post("/clear-data", (_req, res) => {
 function clearAll() {
   db.prepare("DELETE FROM token_usage").run();
   db.prepare("DELETE FROM token_usage_hourly").run();
+  forgetHourly();
   db.prepare("DELETE FROM events").run();
   db.prepare("DELETE FROM agents").run();
   db.prepare("DELETE FROM sessions").run();
@@ -782,6 +784,7 @@ router.post("/cleanup", (req, res) => {
         db.prepare(`DELETE FROM token_usage_hourly WHERE session_id IN (${placeholders})`).run(
           ...chunk
         );
+        forgetHourly(chunk);
         db.prepare(`DELETE FROM sessions WHERE id IN (${placeholders})`).run(...chunk);
       }
       result.purged_sessions = ids.length;