diff --git a/run_agent.py b/run_agent.py
index f25c94f17a9..6e9877a1182 100644
--- a/run_agent.py
+++ b/run_agent.py
@@ -3601,15 +3601,17 @@ class AIAgent:
         ``reasoning_content`` on every assistant tool-call message; omitting
         it causes the next replay to fail with HTTP 400.
 
-        Also detects Kimi models served through third-party providers (e.g.
-        ollama-cloud) by matching ``kimi`` in the model name.
+        Detection is host-driven, not model-name-driven: aggregators like
+        OpenRouter that re-export Kimi/Moonshot models speak their own
+        protocol and reject ``reasoning_content`` echoes. We only enable the
+        kimi-reasoning replay when the request actually targets a
+        kimi/moonshot endpoint or the dedicated kimi-coding provider.
         """
         return (
             self.provider in {"kimi-coding", "kimi-coding-cn"}
             or base_url_host_matches(self.base_url, "api.kimi.com")
             or base_url_host_matches(self.base_url, "moonshot.ai")
             or base_url_host_matches(self.base_url, "moonshot.cn")
-            or "kimi" in (self.model or "").lower()
         )
 
     def _needs_deepseek_tool_reasoning(self) -> bool: