diff --git a/strix/config/models.py b/strix/config/models.py
index 6f0162a00..ea5d574c5 100644
--- a/strix/config/models.py
+++ b/strix/config/models.py
@@ -807,7 +807,7 @@ def _install_openrouter_stream_cost_capture() -> None:
             # survives between turns.
             body = super().transform_request(*args, **kwargs)
             agent_id = request_log.current_call_context().agent_id
-            if agent_id:
+            if agent_id and load_settings().llm.openrouter_sticky_sessions:
                 session_id = _OPENROUTER_SESSION_IDS.setdefault(agent_id, str(uuid.uuid4()))
                 body.setdefault("session_id", session_id)
             return body
diff --git a/strix/config/settings.py b/strix/config/settings.py
index f6c85dd03..69e8575c0 100644
--- a/strix/config/settings.py
+++ b/strix/config/settings.py
@@ -62,6 +62,10 @@ class LlmSettings(BaseSettings):
     # can read back up to a block short. 128 covers the largest common size
     # (OpenAI; DeepSeek and GLM use 64, vLLM defaults to 16).
     cache_block_tokens: int = Field(default=128, ge=1, alias="STRIX_CACHE_BLOCK_TOKENS")
+    openrouter_sticky_sessions: bool = Field(
+        default=True,
+        alias="STRIX_OPENROUTER_STICKY_SESSIONS",
+    )
     disable_streaming: bool = Field(
         default=False,
         alias="LLM_DISABLE_STREAMING",
diff --git a/tests/test_cost_tracking.py b/tests/test_cost_tracking.py
index 9d7ea6db4..5b108abde 100644
--- a/tests/test_cost_tracking.py
+++ b/tests/test_cost_tracking.py
@@ -384,5 +384,8 @@ def test_openrouter_request_carries_agent_session_id() -> None:
         session_id = body()["session_id"]
         assert str(uuid.UUID(session_id)) == session_id
         assert body()["session_id"] == session_id
+        with patch("strix.config.models.load_settings") as settings:
+            settings.return_value.llm.openrouter_sticky_sessions = False
+            assert "session_id" not in body()
     finally:
         request_log.reset_call_context(token)