diff --git a/run_agent.py b/run_agent.py
index 4c0d3be4b..36452bc68 100644
--- a/run_agent.py
+++ b/run_agent.py
@@ -1748,10 +1748,25 @@ class AIAgent:
 
             aux_base_url = str(getattr(client, "base_url", ""))
             aux_api_key = str(getattr(client, "api_key", ""))
+
+            # Read user-configured context_length for the compression model.
+            # Custom endpoints often don't support /models API queries so
+            # get_model_context_length() falls through to the 128K default,
+            # ignoring the explicit config value.  Pass it as the highest-
+            # priority hint so the configured value is always respected.
+            _aux_cfg = (self.config or {}).get("auxiliary", {}).get("compression", {})
+            _aux_context_config = _aux_cfg.get("context_length") if isinstance(_aux_cfg, dict) else None
+            if _aux_context_config is not None:
+                try:
+                    _aux_context_config = int(_aux_context_config)
+                except (TypeError, ValueError):
+                    _aux_context_config = None
+
             aux_context = get_model_context_length(
                 aux_model,
                 base_url=aux_base_url,
                 api_key=aux_api_key,
+                config_context_length=_aux_context_config,
             )
 
             threshold = self.context_compressor.threshold_tokens