diff --git a/dataflow/serving/__init__.py b/dataflow/serving/__init__.py
index df759981..e91abf95 100644
--- a/dataflow/serving/__init__.py
+++ b/dataflow/serving/__init__.py
@@ -1,4 +1,5 @@
 from .api_llm_serving_request import APILLMServing_request
+from .atlascloud_llm_serving import AtlasCloudLLMServing
 from .local_model_llm_serving import LocalModelLLMServing_vllm
 from .local_model_llm_serving import LocalModelLLMServing_sglang
 from .api_vlm_serving_openai import APIVLMServing_openai
@@ -18,6 +19,7 @@
 __all__ = [
     "APIGoogleVertexAIServing",
     "APILLMServing_request",
+    "AtlasCloudLLMServing",
     "LocalModelLLMServing_vllm",
     "LocalModelLLMServing_sglang",
     "APIVLMServing_openai",
diff --git a/dataflow/serving/atlascloud_llm_serving.py b/dataflow/serving/atlascloud_llm_serving.py
new file mode 100644
index 00000000..e4c220e3
--- /dev/null
+++ b/dataflow/serving/atlascloud_llm_serving.py
@@ -0,0 +1,24 @@
+from .api_llm_serving_request import APILLMServing_request
+
+
+class AtlasCloudLLMServing(APILLMServing_request):
+    """Atlas Cloud OpenAI-compatible LLM serving preset.
+
+    Uses ``ATLASCLOUD_API_KEY`` by default and points to the Atlas Cloud
+    chat-completions endpoint while keeping the same request/response behavior
+    as ``APILLMServing_request``.
+    """
+
+    def __init__(
+        self,
+        api_url: str = "https://api.atlascloud.ai/v1/chat/completions",
+        key_name_of_api_key: str = "ATLASCLOUD_API_KEY",
+        model_name: str = "qwen/qwen3.5-flash",
+        **configs,
+    ):
+        super().__init__(
+            api_url=api_url,
+            key_name_of_api_key=key_name_of_api_key,
+            model_name=model_name,
+            **configs,
+        )
diff --git a/test/cpu_only/test_atlascloud_llm_serving.py b/test/cpu_only/test_atlascloud_llm_serving.py
new file mode 100644
index 00000000..62c17fcf
--- /dev/null
+++ b/test/cpu_only/test_atlascloud_llm_serving.py
@@ -0,0 +1,30 @@
+import pytest
+
+from dataflow.serving import AtlasCloudLLMServing
+
+
+def test_atlascloud_defaults(monkeypatch):
+    monkeypatch.setenv("ATLASCLOUD_API_KEY", "dummy-key")
+
+    serving = AtlasCloudLLMServing(max_workers=1, max_retries=1)
+
+    assert serving.api_url == "https://api.atlascloud.ai/v1/chat/completions"
+    assert serving.model_name == "qwen/qwen3.5-flash"
+    assert serving.api_key == "dummy-key"
+    serving.cleanup()
+
+
+def test_atlascloud_uses_custom_model(monkeypatch):
+    monkeypatch.setenv("ATLASCLOUD_API_KEY", "dummy-key")
+
+    serving = AtlasCloudLLMServing(model_name="deepseek-ai/deepseek-v4-pro")
+
+    assert serving.model_name == "deepseek-ai/deepseek-v4-pro"
+    serving.cleanup()
+
+
+def test_atlascloud_requires_api_key(monkeypatch):
+    monkeypatch.delenv("ATLASCLOUD_API_KEY", raising=False)
+
+    with pytest.raises(ValueError, match="ATLASCLOUD_API_KEY"):
+        AtlasCloudLLMServing()