diff --git a/veadk/agent.py b/veadk/agent.py
index e67344aa..e57fa312 100644
--- a/veadk/agent.py
+++ b/veadk/agent.py
@@ -17,6 +17,8 @@
 import os
 from typing import Dict, Literal, Optional, Union
 
+from google.adk.flows.llm_flows.base_llm_flow import BaseLlmFlow
+
 # If user didn't set LITELLM_LOCAL_MODEL_COST_MAP, set it to True
 # to enable local model cost map.
 # This value is `false` by default, which brings heavy performance burden,
@@ -150,6 +152,8 @@ class Agent(LlmAgent):
 
     example_store: Optional[BaseExampleProvider] = None
 
+    enable_supervisor: bool = False
+
     def model_post_init(self, __context: Any) -> None:
         super().model_post_init(None)  # for sub_agents init
 
@@ -402,6 +406,31 @@ def _prepare_tracers(self):
             f"Opentelemetry Tracer init {len(self.tracers[0].exporters)} exporters"  # type: ignore
         )
 
+    @property
+    def _llm_flow(self) -> BaseLlmFlow:
+        from google.adk.flows.llm_flows.auto_flow import AutoFlow
+        from google.adk.flows.llm_flows.single_flow import SingleFlow
+
+        if (
+            self.disallow_transfer_to_parent
+            and self.disallow_transfer_to_peers
+            and not self.sub_agents
+        ):
+            from veadk.flows.supervise_single_flow import SupervisorSingleFlow
+
+            if self.enable_supervisor:
+                logger.debug(f"Enable supervisor flow for agent: {self.name}")
+                return SupervisorSingleFlow(supervised_agent=self)
+            else:
+                return SingleFlow()
+        else:
+            from veadk.flows.supervise_auto_flow import SupervisorAutoFlow
+
+            if self.enable_supervisor:
+                logger.debug(f"Enable supervisor flow for agent: {self.name}")
+                return SupervisorAutoFlow(supervised_agent=self)
+            return AutoFlow()
+
     async def run(self, **kwargs):
         raise NotImplementedError(
             "Run method in VeADK agent is deprecated since version 0.5.6. Please use runner.run_async instead. Ref: https://agentkit.gitbook.io/docs/runner/overview"
diff --git a/veadk/agents/supervise_agent.py b/veadk/agents/supervise_agent.py
new file mode 100644
index 00000000..e2c344ab
--- /dev/null
+++ b/veadk/agents/supervise_agent.py
@@ -0,0 +1,70 @@
+# Copyright (c) 2025 Beijing Volcano Engine Technology Co., Ltd. and/or its affiliates.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from google.adk.models.llm_request import LlmRequest
+from jinja2 import Template
+
+from veadk import Agent, Runner
+from veadk.utils.logger import get_logger
+
+logger = get_logger(__name__)
+
+instruction = Template("""You are a supervisor of an agent system. The system prompt of worker agent is:
+
+```system prompt
+{{ system_prompt }}
+```
+
+
+
+You should guide the agent to finish task and must output a JSON-format string with specific advice and reason:
+                       
+- If you think the history execution is not correct, you should give your advice to the worker agent: {"advice": "Your advice here", "reason": "Your reason here"}.
+- If you think the history execution is correct, you should output an empty string: {"advice": "", "reason": "Your reason here"}.
+""")
+
+
+def build_supervisor(supervised_agent: Agent) -> Agent:
+    custom_instruction = instruction.render(system_prompt=supervised_agent.instruction)
+    agent = Agent(
+        name="supervisor",
+        description="A supervisor for agent execution",
+        instruction=custom_instruction,
+    )
+
+    return agent
+
+
+async def generate_advice(agent: Agent, llm_request: LlmRequest) -> str:
+    runner = Runner(agent=agent)
+
+    messages = ""
+    for content in llm_request.contents:
+        if content and content.parts:
+            for part in content.parts:
+                if part.text:
+                    messages += f"{content.role}: {part.text}"
+                if part.function_call:
+                    messages += f"{content.role}: {part.function_call}"
+                if part.function_response:
+                    messages += f"{content.role}: {part.function_response}"
+
+    prompt = (
+        f"Agent has the following tools: {llm_request.tools_dict}. History trajectory is: "
+        + messages
+    )
+
+    logger.debug(f"Prompt for supervisor: {prompt}")
+
+    return await runner.run(messages=prompt)
diff --git a/veadk/flows/__init__.py b/veadk/flows/__init__.py
new file mode 100644
index 00000000..7f463206
--- /dev/null
+++ b/veadk/flows/__init__.py
@@ -0,0 +1,13 @@
+# Copyright (c) 2025 Beijing Volcano Engine Technology Co., Ltd. and/or its affiliates.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
diff --git a/veadk/flows/supervise_auto_flow.py b/veadk/flows/supervise_auto_flow.py
new file mode 100644
index 00000000..41e16e1a
--- /dev/null
+++ b/veadk/flows/supervise_auto_flow.py
@@ -0,0 +1,70 @@
+# Copyright (c) 2025 Beijing Volcano Engine Technology Co., Ltd. and/or its affiliates.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import json
+from typing import AsyncGenerator
+
+from google.adk.agents.invocation_context import InvocationContext
+from google.adk.events import Event
+from google.adk.models.llm_request import LlmRequest
+from google.adk.models.llm_response import LlmResponse
+from google.genai.types import Content, Part
+
+from veadk import Agent
+from veadk.agents.supervise_agent import generate_advice
+from veadk.flows.supervise_single_flow import SupervisorSingleFlow
+from veadk.utils.logger import get_logger
+
+logger = get_logger(__name__)
+
+
+class SupervisorAutoFlow(SupervisorSingleFlow):
+    def __init__(self, supervised_agent: Agent):
+        super().__init__(supervised_agent)
+
+    async def _call_llm_async(
+        self,
+        invocation_context: InvocationContext,
+        llm_request: LlmRequest,
+        model_response_event: Event,
+    ) -> AsyncGenerator[LlmResponse, None]:
+        supervisor_response = await generate_advice(self._supervisor, llm_request)
+        logger.debug(f"Advice from supervisor: {supervisor_response}")
+
+        advice_and_reason = json.loads(supervisor_response)
+
+        if advice_and_reason["advice"]:
+            logger.debug("Add supervisor advice to llm request.")
+            llm_request.contents.append(
+                Content(
+                    parts=[
+                        Part(
+                            text=f"""Message from your supervisor (not user): {advice_and_reason["advice"]}, the corresponding reason is {advice_and_reason["reason"]}
+
+    Please follow the advice and reason above to optimize your actions.
+    """
+                        )
+                    ],
+                    role="user",
+                )
+            )
+        else:
+            logger.info(
+                f"Supervisor advice is empty, reason: {advice_and_reason['reason']}. Skip adding to llm request."
+            )
+
+        async for llm_response in super()._call_llm_async(
+            invocation_context, llm_request, model_response_event
+        ):
+            yield llm_response
diff --git a/veadk/flows/supervise_single_flow.py b/veadk/flows/supervise_single_flow.py
new file mode 100644
index 00000000..fc8de1b6
--- /dev/null
+++ b/veadk/flows/supervise_single_flow.py
@@ -0,0 +1,44 @@
+# Copyright (c) 2025 Beijing Volcano Engine Technology Co., Ltd. and/or its affiliates.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import AsyncGenerator
+
+from google.adk.agents.invocation_context import InvocationContext
+from google.adk.events import Event
+from google.adk.flows.llm_flows.single_flow import SingleFlow
+from google.adk.models.llm_request import LlmRequest
+from google.adk.models.llm_response import LlmResponse
+from typing_extensions import override
+
+from veadk import Agent
+from veadk.agents.supervise_agent import build_supervisor
+
+
+class SupervisorSingleFlow(SingleFlow):
+    def __init__(self, supervised_agent: Agent):
+        self._supervisor = build_supervisor(supervised_agent)
+
+        super().__init__()
+
+    @override
+    async def _call_llm_async(
+        self,
+        invocation_context: InvocationContext,
+        llm_request: LlmRequest,
+        model_response_event: Event,
+    ) -> AsyncGenerator[LlmResponse, None]:
+        async for llm_response in super()._call_llm_async(
+            invocation_context, llm_request, model_response_event
+        ):
+            yield llm_response