diff --git a/docs/source/guides/clients.rst b/docs/source/guides/clients.rst
index 3adf456df19..55702f14d2d 100644
--- a/docs/source/guides/clients.rst
+++ b/docs/source/guides/clients.rst
@@ -133,6 +133,50 @@ Alternatively, use the ``server_ready_timeout`` parameter to specify the maximum
 
     client.close()
 
+.. _call-a-task-endpoint:
+
+Call a task endpoint
+--------------------
+
+You can create clients to interact with Services defined with :ref:`task <bentoml-tasks>` endpoints by submitting inputs and then asynchronously checking for results at a later time. This model is particularly useful for “fire-and-forget" operations where the client does not need to actively wait for the task to complete. Both synchronous (``SyncHTTPClient``) and asynchronous (``AsyncHTTPClient``) clients can be used to call a task endpoint.
+
+To submit a task:
+
+.. code-block:: python
+
+    import bentoml
+    import numpy as np
+
+    input_data = np.array([...])  # Some numpy array
+
+    client = bentoml.SyncHTTPClient('http://localhost:3000')
+    # The arguments are the same as the Service method, just call with `.submit()`
+    task = client.long_running_task.submit(input=input_data)
+    print("Task submitted, ID:", task.id)
+
+Check the task status and retrieve the result at a later time:
+
+.. code-block:: python
+
+    status = task.get_status()
+    if status.value == 'success':
+        print("The task runs successfully. The result is", task.get())
+    elif status.value == 'failure':
+        print("The task run failed.")
+    else:
+        print("The task is still running.")
+
+Use ``retry()`` if a task fails or you need to rerun the task with the same parameters:
+
+.. code-block:: python
+
+    status = task.get_status()
+    if status.value == 'failure':
+        print("Task failed, retrying...")
+        new_task = task.retry()
+        new_status = new_task.get_status()
+        print("New task status:", new_status.value)
+
 Input and output
 ----------------
 
diff --git a/docs/source/guides/services.rst b/docs/source/guides/services.rst
index 038f7bab335..b8c97fa904e 100644
--- a/docs/source/guides/services.rst
+++ b/docs/source/guides/services.rst
@@ -140,7 +140,7 @@ BentoML's lifecycle hooks provide a way to insert custom logic at specific stage
 You use decorators to set lifecycle hooks. For details, see :doc:`/guides/lifecycle-hooks`.
 
 Synchronous and asynchronous APIs
-----------------------------------
+---------------------------------
 
 APIs in a BentoML Service can be defined as either synchronous functions or asynchronous coroutines in Python.
 
@@ -259,6 +259,30 @@ However, directly calling synchronous blocking functions within an asynchronous
 
 In this example, the ``.to_async`` property converts synchronous methods (``txt2img`` and ``synthesize`` of ``SDXLTurboService`` and ``XTTSService`` respectively) into their asynchronous versions, enabling the ``generate_card`` method to perform multiple asynchronous operations concurrently with ``asyncio.gather``.
 
+.. _bentoml-tasks:
+
+Tasks
+-----
+
+Tasks in BentoML allow you to execute long-running operations in the background, managed via a task queue style API. These background tasks are ideal for scenarios like batch processing and image or video generation where you don't need the results immediately or synchronously.
+
+To define a task endpoint, use the ``@bentoml.task`` decorator in the Service constructor. Here's an example:
+
+.. code-block:: python
+
+    import bentoml
+    import numpy as np
+
+    @bentoml.service
+    class MyService:
+        @bentoml.task
+        def long_running_task(self, input: np.ndarray) -> np.ndarray:
+            # Simulated long-running process
+            # Placeholder for actual processing logic
+            return result
+
+BentoML automatically exposes several endpoints for clients to manage the task, such as task submission, status retrieval, and cancellation. For more information, see :ref:`call-a-task-endpoint`.
+
 Convert legacy Runners to a Service
 -----------------------------------