diff --git a/examples/realtime_stt.py b/examples/realtime_stt.py
index c46c187..35886d7 100644
--- a/examples/realtime_stt.py
+++ b/examples/realtime_stt.py
@@ -20,7 +20,7 @@ def on_mic(indata, frames, t, status):
         except queue.Full:
             pass
 
-    palabra = Palabra() # set your credentials here or vie ENV
+    palabra = Palabra()  # set your credentials here or vie ENV
 
     async with palabra.stt(language="ru", translate_languages=["es", "en"]) as stt:
 
diff --git a/examples/realtime_tts.py b/examples/realtime_tts.py
index 48e609e..b2aed02 100644
--- a/examples/realtime_tts.py
+++ b/examples/realtime_tts.py
@@ -4,7 +4,7 @@
 
 
 async def main():
-    palabra = Palabra() # set your credentials here or vie ENV
+    palabra = Palabra()  # set your credentials here or vie ENV
 
     async with palabra.tts(language="en", voice_id="default_low") as tts:
         # one-shot: send text, collect all chunks
diff --git a/examples/sts_buffer_streaming.py b/examples/sts_buffer_streaming.py
index 2cc8776..b20ebf3 100644
--- a/examples/sts_buffer_streaming.py
+++ b/examples/sts_buffer_streaming.py
@@ -25,7 +25,7 @@ async def feed_source():
 
 
 async def main():
-    palabra = Palabra() # set your credentials here or vie ENV
+    palabra = Palabra()  # set your credentials here or vie ENV
 
     async with palabra.translation(source="en", targets=["es"]) as session:
 
diff --git a/examples/sts_file_to_file.py b/examples/sts_file_to_file.py
index d1c88ad..efdf147 100644
--- a/examples/sts_file_to_file.py
+++ b/examples/sts_file_to_file.py
@@ -1,6 +1,6 @@
 from palabra_ai import Palabra
 
-palabra = Palabra() # set your credentials here or vie ENV
+palabra = Palabra()  # set your credentials here or vie ENV
 
 palabra.translate_file(
     "speech_en.wav",
diff --git a/examples/sts_mic_to_speakers.py b/examples/sts_mic_to_speakers.py
index d6314dc..498dba7 100644
--- a/examples/sts_mic_to_speakers.py
+++ b/examples/sts_mic_to_speakers.py
@@ -35,7 +35,7 @@ def on_spk(outdata, frames, t, status):
         else:
             outdata.fill(0)
 
-    palabra = Palabra() # set your credentials here or vie ENV
+    palabra = Palabra()  # set your credentials here or vie ENV
     async with palabra.translation(source="en", targets=["es"]) as session:
 
         async def feed():
diff --git a/examples/sts_multi_language.py b/examples/sts_multi_language.py
index 8c972ce..27eae90 100644
--- a/examples/sts_multi_language.py
+++ b/examples/sts_multi_language.py
@@ -1,6 +1,6 @@
 from palabra_ai import Palabra
 
-palabra = Palabra() # set your credentials here or vie ENV
+palabra = Palabra()  # set your credentials here or vie ENV
 
 results = palabra.translate_file(
     "presentation.mp3",  # mp3 needs: pip install palabra-ai[audio]
diff --git a/pyproject.toml b/pyproject.toml
index 811dc34..fff8d8a 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "palabra-ai"
-version = "1.0.2"
+version = "1.0.3"
 description = "Simple Python client for Palabra AI APIs"
 readme = "README.md"
 requires-python = ">=3.10"
diff --git a/src/palabra_ai/client.py b/src/palabra_ai/client.py
index 9ecd562..49ed850 100644
--- a/src/palabra_ai/client.py
+++ b/src/palabra_ai/client.py
@@ -27,6 +27,7 @@
 GET_TASK_INTERVAL = 2.1  # server allows 1 get_task per 2s
 SESSION_RETRIES = 3  # create_session attempts on network errors / 5xx
 RETRY_BACKOFF = 0.5  # seconds; doubles per attempt (0.5, 1.0)
+S2S_SESSION_INTENT = "api"
 
 
 @dataclass(frozen=True)
@@ -69,12 +70,19 @@ def _headers(self) -> dict[str, str]:
             )
         return {"ClientID": self.client_id, "ClientSecret": self.client_secret}
 
-    async def create_session(self) -> Session:
+    async def create_session(self, *, intent: str | None = None) -> Session:
         """Create a streaming session via REST.
 
+        intent is the session kind sent as data.intent; session-storage routes
+        and bills by it. Each product hardcodes its own value (s2s "api",
+        tts "tts_api", stt "stt") — not a user choice. Omitted when None.
+
         Transient failures (network errors, 5xx) are retried up to
         SESSION_RETRIES times with backoff; 4xx fails immediately.
         """
+        data: dict[str, Any] = {}
+        if intent is not None:
+            data["intent"] = intent
         last_error: Exception | None = None
         for attempt in range(SESSION_RETRIES):
             if attempt:
@@ -84,9 +92,7 @@ async def create_session(self) -> Session:
                     resp = await client.post(
                         f"{self.api_url}/session-storage/session",
                         headers=self._headers(),
-                        json={
-                            "data": {}
-                        },
+                        json={"data": data},
                     )
             except httpx.TransportError as e:
                 last_error = e
@@ -332,7 +338,7 @@ def task(self) -> dict[str, Any]:
 
     async def __aenter__(self) -> TranslationSession:
         if self._session is None:
-            self._session = await self._palabra.create_session()
+            self._session = await self._palabra.create_session(intent=S2S_SESSION_INTENT)
         url = f"{self._session.ws_url}?token={self._session.publisher}"
         try:
             self._ws = await websockets.connect(url, ping_interval=10, ping_timeout=30, max_size=None)
diff --git a/src/palabra_ai/stt.py b/src/palabra_ai/stt.py
index 93f1f1c..a524448 100644
--- a/src/palabra_ai/stt.py
+++ b/src/palabra_ai/stt.py
@@ -18,6 +18,7 @@
     from .client import Palabra, Session
 
 ASR_STREAM_PATH = "/asr/v1/speech-to-text/stream"
+SESSION_INTENT = "stt"  # session kind for billing routing
 DEFAULT_SAMPLE_RATE = 16000  # ASR recommended/default input rate (translation/TTS use 24k)
 
 _STT_TRANSCRIPT_TYPES = frozenset({"transcription", "translated_transcription"})
@@ -110,7 +111,7 @@ async def __aenter__(self) -> SttSession:
             base, token = self._direct
         else:
             if self._session is None:
-                self._session = await self._palabra.create_session()
+                self._session = await self._palabra.create_session(intent=SESSION_INTENT)
             base = _asr_ws_url(self._palabra.api_url)
             token = self._session.publisher
         url = f"{base}?{urlencode({'token': token, **self._params}, safe=',')}"
diff --git a/src/palabra_ai/tts.py b/src/palabra_ai/tts.py
index a457c50..dc1d797 100644
--- a/src/palabra_ai/tts.py
+++ b/src/palabra_ai/tts.py
@@ -20,6 +20,7 @@
 
 # The Realtime TTS endpoint is fixed (not taken from the session response).
 TTS_STREAM_URL = "wss://stream.palabra.ai/tts-api/v1/text-to-speech/stream"
+SESSION_INTENT = "tts_api"  # session kind for billing routing
 
 
 @dataclass(frozen=True)
@@ -80,7 +81,7 @@ async def __aenter__(self) -> TtsSession:
             url = f"{self._direct[0]}?token={self._direct[1]}"
         else:
             if self._session is None:
-                self._session = await self._palabra.create_session()
+                self._session = await self._palabra.create_session(intent=SESSION_INTENT)
             url = f"{TTS_STREAM_URL}?token={self._session.publisher}"
         try:
             self._ws = await websockets.connect(url, ping_interval=10, ping_timeout=30, max_size=None)
diff --git a/uv.lock b/uv.lock
index 84b39ff..cc7ee00 100644
--- a/uv.lock
+++ b/uv.lock
@@ -391,7 +391,7 @@ wheels = [
 
 [[package]]
 name = "palabra-ai"
-version = "1.0.2"
+version = "1.0.3"
 source = { editable = "." }
 dependencies = [
     { name = "httpx" },