diff --git a/packages/app/src/contexts/realtime-context.tsx b/packages/app/src/contexts/realtime-context.tsx
index c05f8ad84..f1985ae86 100644
--- a/packages/app/src/contexts/realtime-context.tsx
+++ b/packages/app/src/contexts/realtime-context.tsx
@@ -41,6 +41,19 @@ export function RealtimeProvider({ children }: RealtimeProviderProps) {
       if (isPlayingAudio) {
         audioPlayer.stop();
       }
+      // Abort any in-flight LLM turn before the new speech segment streams
+      try {
+        const abortMessage: WSInboundMessage = {
+          type: "session",
+          message: {
+            type: "abort_request",
+          },
+        };
+        ws.send(abortMessage);
+        console.log("[Realtime] Sent abort_request before streaming audio");
+      } catch (error) {
+        console.error("[Realtime] Failed to send abort_request:", error);
+      }
     },
     onSpeechEnd: () => {
       console.log("[Realtime] Speech ended");
diff --git a/plan.md b/plan.md
index 2fc4043a1..1b3b8dac4 100644
--- a/plan.md
+++ b/plan.md
@@ -54,7 +54,8 @@
 
 ## Voice Interruptions & Streaming
 
-- [ ] Wire interrupt signal on speech start so VAD immediately issues `abort_request` to the server before any audio segment is sent.
+- [x] Wire interrupt signal on speech start so VAD immediately issues `abort_request` to the server before any audio segment is sent.
+  - Realtime speech detection now sends a `session/abort_request` as soon as VAD confirms speech, ensuring any in-flight LLM turn is interrupted before the buffered audio is uploaded (implemented in `packages/app/src/contexts/realtime-context.tsx` with logging/error handling).
 - [ ] Abort server-side playback/LLM as soon as a realtime audio chunk arrives; set a speech-in-progress flag that pauses new TTS until the turn completes.
 - [ ] Interrupt the currently focused agent whenever a realtime user turn begins, ensuring the next prompt starts fresh.
 - [ ] Harden playback stop/queue clearing so speech detection purges suppressed audio and the server stops emitting TTS while the user is talking.