diff --git a/sdk/agentserver/azure-ai-agentserver-responses/CHANGELOG.md b/sdk/agentserver/azure-ai-agentserver-responses/CHANGELOG.md
index 312ac5bb7990..4ae7f4fa224b 100644
--- a/sdk/agentserver/azure-ai-agentserver-responses/CHANGELOG.md
+++ b/sdk/agentserver/azure-ai-agentserver-responses/CHANGELOG.md
@@ -1,6 +1,10 @@
 # Release History
 
-## 2.1.0b1 (2026-08-11)
+## 2.1.0b1 (2026-08-13)
+
+### Bugs Fixed
+
+- `get_history_item_ids` on the in-memory and file stores now keeps the newest item IDs when applying `limit`. (#48514)
 
 ### Breaking Changes
 
diff --git a/sdk/agentserver/azure-ai-agentserver-responses/azure/ai/agentserver/responses/store/_file.py b/sdk/agentserver/azure-ai-agentserver-responses/azure/ai/agentserver/responses/store/_file.py
index 1ff434cb118a..6069faea9044 100644
--- a/sdk/agentserver/azure-ai-agentserver-responses/azure/ai/agentserver/responses/store/_file.py
+++ b/sdk/agentserver/azure-ai-agentserver-responses/azure/ai/agentserver/responses/store/_file.py
@@ -579,7 +579,8 @@ async def get_history_item_ids(
 
             if limit <= 0:
                 return []
-            return resolved[:limit]
+            # Keep the most recent N history items (drop oldest when over limit).
+            return resolved[-limit:]
 
     # ------------------------------------------------------------------
     # Internal helpers (must be called with self._lock held)
diff --git a/sdk/agentserver/azure-ai-agentserver-responses/azure/ai/agentserver/responses/store/_memory.py b/sdk/agentserver/azure-ai-agentserver-responses/azure/ai/agentserver/responses/store/_memory.py
index 941be36ee682..97dec94f1d2f 100644
--- a/sdk/agentserver/azure-ai-agentserver-responses/azure/ai/agentserver/responses/store/_memory.py
+++ b/sdk/agentserver/azure-ai-agentserver-responses/azure/ai/agentserver/responses/store/_memory.py
@@ -325,7 +325,8 @@ async def get_history_item_ids(
 
             if limit <= 0:
                 return []
-            return resolved[:limit]
+            # Keep the most recent N history items (drop oldest when over limit).
+            return resolved[-limit:]
 
     async def create_execution(self, execution: ResponseExecution, *, ttl_seconds: int | None = None) -> None:
         """Create a new execution and replay container for ``execution.response_id``.
diff --git a/sdk/agentserver/azure-ai-agentserver-responses/tests/unit/test_file_response_store_parity.py b/sdk/agentserver/azure-ai-agentserver-responses/tests/unit/test_file_response_store_parity.py
index 89a94485b6f6..0779ed7cf260 100644
--- a/sdk/agentserver/azure-ai-agentserver-responses/tests/unit/test_file_response_store_parity.py
+++ b/sdk/agentserver/azure-ai-agentserver-responses/tests/unit/test_file_response_store_parity.py
@@ -327,7 +327,8 @@ async def test_history_respects_limit(tmp_path: Path) -> None:
             history_item_ids=["hist1", "hist2"],
         )
         ids = await provider.get_history_item_ids("r_prev", None, limit=3)
-        assert ids == ["hist1", "hist2", "in1"]
+        # Chronological order is oldest-first; truncation must keep the newest IDs.
+        assert ids == ["out1", "out2", "out3"]
         # Non-positive limit returns empty.
         ids_zero = await provider.get_history_item_ids("r_prev", None, limit=0)
         assert ids_zero == []
diff --git a/sdk/agentserver/azure-ai-agentserver-responses/tests/unit/test_in_memory_provider_crud.py b/sdk/agentserver/azure-ai-agentserver-responses/tests/unit/test_in_memory_provider_crud.py
index 3915baeb02fd..9b36e9b0f77a 100644
--- a/sdk/agentserver/azure-ai-agentserver-responses/tests/unit/test_in_memory_provider_crud.py
+++ b/sdk/agentserver/azure-ai-agentserver-responses/tests/unit/test_in_memory_provider_crud.py
@@ -374,6 +374,8 @@ def test_history__respects_limit() -> None:
 
     ids = asyncio.run(provider.get_history_item_ids("resp_lim", None, 3))
     assert len(ids) == 3
+    # Chronological order is oldest-first; truncation must keep the newest IDs.
+    assert ids == ["in_lim_7", "in_lim_8", "in_lim_9"]
 
 
 def test_history__zero_limit_returns_empty() -> None: