diff --git a/hermes_cli/config.py b/hermes_cli/config.py
index 000d8626b..f3c6073a6 100644
--- a/hermes_cli/config.py
+++ b/hermes_cli/config.py
@@ -317,6 +317,8 @@ DEFAULT_CONFIG = {
         "provider": "",    # e.g. "openrouter" (empty = inherit parent provider + credentials)
         "base_url": "",    # direct OpenAI-compatible endpoint for subagents
         "api_key": "",     # API key for delegation.base_url (falls back to OPENAI_API_KEY)
+        "max_iterations": 50,  # per-subagent iteration cap (each subagent gets its own budget,
+                               # independent of the parent's max_iterations)
     },
 
     # Ephemeral prefill messages file — JSON list of {role, content} dicts
diff --git a/run_agent.py b/run_agent.py
index 5ed5d7245..60be56129 100644
--- a/run_agent.py
+++ b/run_agent.py
@@ -162,11 +162,15 @@ def _install_safe_stdio() -> None:
 
 
 class IterationBudget:
-    """Thread-safe shared iteration counter for parent and child agents.
+    """Thread-safe iteration counter for an agent.
 
-    Tracks total LLM-call iterations consumed across a parent agent and all
-    its subagents.  A single ``IterationBudget`` is created by the parent
-    and passed to every child so they share the same cap.
+    Each agent (parent or subagent) gets its own ``IterationBudget``.
+    The parent's budget is capped at ``max_iterations`` (default 90).
+    Each subagent gets an independent budget capped at
+    ``delegation.max_iterations`` (default 50) — this means total
+    iterations across parent + subagents can exceed the parent's cap.
+    Users control the per-subagent limit via ``delegation.max_iterations``
+    in config.yaml.
 
     ``execute_code`` (programmatic tool calling) iterations are refunded via
     :meth:`refund` so they don't eat into the budget.
@@ -5719,7 +5723,7 @@ class AIAgent:
             api_call_count += 1
             if not self.iteration_budget.consume():
                 if not self.quiet_mode:
-                    self._safe_print(f"\n⚠️  Session iteration budget exhausted ({self.iteration_budget.max_total} total across agent + subagents)")
+                    self._safe_print(f"\n⚠️  Iteration budget exhausted ({self.iteration_budget.used}/{self.iteration_budget.max_total} iterations used)")
                 break
 
             # Fire step_callback for gateway hooks (agent:step event)
@@ -7178,7 +7182,7 @@ class AIAgent:
             or self.iteration_budget.remaining <= 0
         ):
             if self.iteration_budget.remaining <= 0 and not self.quiet_mode:
-                print(f"\n⚠️  Session iteration budget exhausted ({self.iteration_budget.used}/{self.iteration_budget.max_total} used, including subagents)")
+                print(f"\n⚠️  Iteration budget exhausted ({self.iteration_budget.used}/{self.iteration_budget.max_total} iterations used)")
             final_response = self._handle_max_iterations(messages, api_call_count)
         
         # Determine if conversation completed successfully
diff --git a/tools/delegate_tool.py b/tools/delegate_tool.py
index 36c6dad98..57f2a4bee 100644
--- a/tools/delegate_tool.py
+++ b/tools/delegate_tool.py
@@ -191,9 +191,10 @@ def _build_child_agent(
     # Build progress callback to relay tool calls to parent display
     child_progress_cb = _build_child_progress_callback(task_index, parent_agent)
 
-    # Share the parent's iteration budget so subagent tool calls
-    # count toward the session-wide limit.
-    shared_budget = getattr(parent_agent, "iteration_budget", None)
+    # Each subagent gets its own iteration budget capped at max_iterations
+    # (configurable via delegation.max_iterations, default 50).  This means
+    # total iterations across parent + subagents can exceed the parent's
+    # max_iterations.  The user controls the per-subagent cap in config.yaml.
 
     # Resolve effective credentials: config override > parent inherit
     effective_model = model or parent_agent.model
@@ -230,7 +231,7 @@ def _build_child_agent(
         providers_order=parent_agent.providers_order,
         provider_sort=parent_agent.provider_sort,
         tool_progress_callback=child_progress_cb,
-        iteration_budget=shared_budget,
+        iteration_budget=None,  # fresh budget per subagent
     )
     # Set delegation depth so children can't spawn grandchildren
     child._delegate_depth = getattr(parent_agent, '_delegate_depth', 0) + 1