diff --git a/frontend/src/app/pages/AgentChat/bubbles/StreamingBubble.tsx b/frontend/src/app/pages/AgentChat/bubbles/StreamingBubble.tsx
index f4b4c0c5..fe6659b1 100644
--- a/frontend/src/app/pages/AgentChat/bubbles/StreamingBubble.tsx
+++ b/frontend/src/app/pages/AgentChat/bubbles/StreamingBubble.tsx
@@ -2,6 +2,7 @@ import React, { useEffect, useRef } from 'react';
 import { useStreamingMessage } from '@/shared/state/streamingSlice';
 import MessageBubble from './MessageBubble';
 import ToolCallBubble from '../tool-bubbles/ToolCallBubble';
+import { useSmoothText } from './useSmoothText';
 
 interface Props {
   sessionId: string;
@@ -13,7 +14,15 @@ interface Props {
 /** Leaf subscriber for one session's streaming entry; isolates re-renders so AgentChat doesn't churn per character. */
 const StreamingBubble: React.FC<Props> = ({ sessionId, activeBranchId, turnLabel, onStreamGrew }) => {
   const streamingMessage = useStreamingMessage(sessionId);
-  const typedContent = streamingMessage?.content ?? '';
+  const rawContent = streamingMessage?.content ?? '';
+  // Smooth-reveal the assistant's generated text at a steady cadence so it reads
+  // like typing instead of bursty network chunks. Provider-agnostic by design:
+  // every model (Anthropic/OpenAI/Gemini/OpenRouter/custom) funnels through this
+  // same streaming slice, so smoothing here covers all of them at once. Tool-call
+  // input is left raw (it's args, not prose). Zero added TTFT (see useSmoothText).
+  const isTextRole = streamingMessage?.role !== 'tool_call';
+  const smoothContent = useSmoothText(rawContent, isTextRole);
+  const typedContent = isTextRole ? smoothContent : rawContent;
   // RAF-coalesce so onStreamGrew fires once per frame regardless of token rate.
   const onGrewRef = useRef(onStreamGrew);
   onGrewRef.current = onStreamGrew;
diff --git a/frontend/src/app/pages/AgentChat/bubbles/useSmoothText.ts b/frontend/src/app/pages/AgentChat/bubbles/useSmoothText.ts
new file mode 100644
index 00000000..b3d603bd
--- /dev/null
+++ b/frontend/src/app/pages/AgentChat/bubbles/useSmoothText.ts
@@ -0,0 +1,75 @@
+import { useEffect, useRef, useState } from 'react';
+
+/**
+ * Smoothly reveals streamed text at a steady cadence instead of painting bursty
+ * network chunks as they land. Decouples DISPLAY rate from ARRIVAL rate the way
+ * claude.ai does, so generated text reads like it's being typed rather than
+ * dumped in clumps.
+ *
+ * Zero dependencies. Zero added TTFT: the first characters reveal on the very next
+ * animation frame after the first delta (same frame budget as painting it directly).
+ * The reveal rate is ADAPTIVE — it accelerates as the backlog grows, so display
+ * never falls meaningfully behind the model and never reads as laggy. The rAF loop
+ * runs ONLY while there's a backlog to drain and parks itself at zero cost once
+ * caught up, so it adds no idle-frame churn.
+ */
+
+/** Pure pacing step (exported for testing): chars to reveal this frame. */
+export function smoothStep(shown: number, full: number): number {
+  if (shown >= full) return full;
+  const backlog = full - shown;
+  // Floor of 3 chars/frame (~180 chars/sec at 60fps) for a calm typing feel,
+  // and drain ~1/4 of any backlog on top of that so bursts catch up fast. The
+  // /4 keeps mid-stream lag small (a few words at most), so when the live bubble
+  // hands off to the final message at stream end there's no visible jump. Never
+  // overshoots `full`.
+  const step = Math.max(3, Math.ceil(backlog / 4));
+  return Math.min(full, shown + step);
+}
+
+export function useSmoothText(target: string, enabled: boolean): string {
+  const [shownLen, setShownLen] = useState(enabled ? 0 : target.length);
+  const rafRef = useRef<number | null>(null);
+  const targetRef = useRef(target);
+  targetRef.current = target;
+
+  useEffect(() => {
+    // Disabled (historical message, or smoothing turned off): show all, stop loop.
+    if (!enabled) {
+      if (rafRef.current != null) { cancelAnimationFrame(rafRef.current); rafRef.current = null; }
+      setShownLen(targetRef.current.length);
+      return;
+    }
+    const tick = () => {
+      rafRef.current = null;
+      setShownLen((cur) => {
+        const next = smoothStep(cur, targetRef.current.length);
+        if (next < targetRef.current.length) rafRef.current = requestAnimationFrame(tick);
+        return next;
+      });
+    };
+    // Start a drain only if we're behind and no loop is already running.
+    if (rafRef.current == null && shownLen < target.length) {
+      rafRef.current = requestAnimationFrame(tick);
+    }
+    return () => {
+      if (rafRef.current != null) { cancelAnimationFrame(rafRef.current); rafRef.current = null; }
+    };
+  }, [enabled, target.length, shownLen]);
+
+  // Target shrank (new turn / reset / branch switch): re-sync so we don't slice
+  // past the end of a shorter string.
+  useEffect(() => {
+    if (shownLen > target.length) setShownLen(enabled ? 0 : target.length);
+  }, [target.length, shownLen, enabled]);
+
+  // ZERO added TTFT: on the very first frame content exists (shownLen still 0),
+  // reveal the floor immediately in-render instead of waiting a frame for the rAF
+  // tick. Pure derivation, no extra render — so first visible text lands on the
+  // exact same frame it would have without smoothing. State catches up next frame.
+  if (!enabled) return target;
+  const effectiveShown = (shownLen === 0 && target.length > 0)
+    ? Math.min(3, target.length)
+    : shownLen;
+  return target.slice(0, effectiveShown);
+}