diff --git a/.changeset/ninety-items-yell.md b/.changeset/ninety-items-yell.md
new file mode 100644
index 000000000..7bea9e8fd
--- /dev/null
+++ b/.changeset/ninety-items-yell.md
@@ -0,0 +1,5 @@
+---
+"livekit-agents": patch
+---
+
+voiceassistant: add VoiceAssistantState
diff --git a/livekit-agents/livekit/agents/voice_assistant/voice_assistant.py b/livekit-agents/livekit/agents/voice_assistant/voice_assistant.py
index 18d0b8cb2..f0a3dd17c 100644
--- a/livekit-agents/livekit/agents/voice_assistant/voice_assistant.py
+++ b/livekit-agents/livekit/agents/voice_assistant/voice_assistant.py
@@ -34,6 +34,8 @@
     "function_calls_finished",
 ]
 
+VoiceAssistantState = Literal["initializing", "listening", "thinking", "speaking"]
+
 
 _CallContextVar = contextvars.ContextVar["AssistantCallContext"](
     "voice_assistant_contextvar"
@@ -201,6 +203,8 @@ def __init__(
 
         self._last_end_of_speech_time: float | None = None
 
+        self._update_state_task: asyncio.Task | None = None
+
     @property
     def fnc_ctx(self) -> FunctionContext | None:
         return self._fnc_ctx
@@ -301,6 +305,23 @@ async def say(
         new_handle.initialize(source=source, synthesis_handle=synthesis_handle)
         self._add_speech_for_playout(new_handle)
 
+    def _update_state(self, state: VoiceAssistantState, delay: float = 0.0):
+        """Set the current state of the voice assistant"""
+
+        @utils.log_exceptions(logger=logger)
+        async def _run_task(delay: float) -> None:
+            await asyncio.sleep(delay)
+
+            if self._room.isconnected():
+                await self._room.local_participant.set_attributes(
+                    {"voice_assistant.state": state}
+                )
+
+        if self._update_state_task is not None:
+            self._update_state_task.cancel()
+
+        self._update_state_task = asyncio.create_task(_run_task(delay))
+
     async def aclose(self) -> None:
         """Close the voice assistant"""
         if not self._started:
@@ -333,6 +354,7 @@ def _on_start_of_speech(ev: vad.VADEvent) -> None:
             self._plotter.plot_event("user_started_speaking")
             self.emit("user_started_speaking")
             self._deferred_validation.on_human_start_of_speech(ev)
+            self._update_state("listening")
 
         def _on_vad_updated(ev: vad.VADEvent) -> None:
             if not self._track_published_fut.done():
@@ -393,6 +415,7 @@ async def _main_task(self) -> None:
         if self._opts.plotting:
             await self._plotter.start()
 
+        self._update_state("initializing")
         audio_source = rtc.AudioSource(self._tts.sample_rate, self._tts.num_channels)
         track = rtc.LocalAudioTrack.create_audio_track("assistant_voice", audio_source)
         self._agent_publication = await self._room.local_participant.publish_track(
@@ -410,10 +433,12 @@ async def _main_task(self) -> None:
         def _on_playout_started() -> None:
             self._plotter.plot_event("agent_started_speaking")
             self.emit("agent_started_speaking")
+            self._update_state("speaking")
 
         def _on_playout_stopped(interrupted: bool) -> None:
             self._plotter.plot_event("agent_stopped_speaking")
             self.emit("agent_stopped_speaking")
+            self._update_state("listening")
 
         agent_playout.on("playout_started", _on_playout_started)
         agent_playout.on("playout_stopped", _on_playout_stopped)
@@ -439,6 +464,9 @@ def _synthesize_agent_reply(self) -> None:
         if self._pending_agent_reply is not None:
             self._pending_agent_reply.interrupt()
 
+        if self._human_input is not None and not self._human_input.speaking:
+            self._update_state("thinking", 0.2)
+
         self._pending_agent_reply = new_handle = SpeechHandle.create_assistant_reply(
             allow_interruptions=self._opts.allow_interruptions,
             add_to_chat_ctx=True,