diff --git a/CMakeLists.txt b/CMakeLists.txt
index af065158699199af61aca02f563dda1b1cddf2b1..7c497e3e048c4dd8d5c1291286de2ab9d218b914 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -85,6 +85,17 @@ include(generic)
 include(flags)
 endif()
 
+if (APP)
+include(external/zlib)
+include(external/boost)
+include(external/protobuf)
+include(external/gflags)
+include(external/glog)
+include(external/pybind11)
+include(external/python)
+include(generic)
+endif()
+
 if (SERVER)
 include(external/cudnn)
 include(paddlepaddle)
diff --git a/core/CMakeLists.txt b/core/CMakeLists.txt
index ce2e5e3814ae1e585976c5d9c8848b506293ee67..56296b53319fb185c772ffa10e8b31c8203862fb 100644
--- a/core/CMakeLists.txt
+++ b/core/CMakeLists.txt
@@ -23,6 +23,11 @@ add_subdirectory(pdcodegen)
 add_subdirectory(sdk-cpp)
 endif()
 
+if (APP)
+add_subdirectory(configure)
+endif()
+
+
 if(CLIENT)
 add_subdirectory(general-client)
 endif()
diff --git a/core/configure/CMakeLists.txt b/core/configure/CMakeLists.txt
index b6384fc99ea3df6d71a61865e3aabf5b39b510dd..d3e5b75da96ad7a0789866a4a2c474fad988c21b 100644
--- a/core/configure/CMakeLists.txt
+++ b/core/configure/CMakeLists.txt
@@ -1,3 +1,4 @@
+if (SERVER OR CLIENT)
 LIST(APPEND protofiles
         ${CMAKE_CURRENT_LIST_DIR}/proto/server_configure.proto
         ${CMAKE_CURRENT_LIST_DIR}/proto/sdk_configure.proto
@@ -28,6 +29,7 @@ FILE(GLOB inc ${CMAKE_CURRENT_BINARY_DIR}/*.pb.h)
 
 install(FILES ${inc}
         DESTINATION ${PADDLE_SERVING_INSTALL_DIR}/include/configure)
+endif()
 
 py_proto_compile(general_model_config_py_proto SRCS proto/general_model_config.proto)
 add_custom_target(general_model_config_py_proto_init ALL COMMAND ${CMAKE_COMMAND} -E touch __init__.py)
@@ -51,6 +53,14 @@ add_custom_command(TARGET general_model_config_py_proto POST_BUILD
 
 endif()
 
+if (APP)
+add_custom_command(TARGET general_model_config_py_proto POST_BUILD
+                COMMAND ${CMAKE_COMMAND} -E make_directory ${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving_app/proto
+                COMMAND cp *.py ${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving_app/proto
+                COMMENT "Copy generated general_model_config proto file into directory paddle_serving_app/proto."
+                WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR})
+endif()
+
 if (SERVER)
 py_proto_compile(server_config_py_proto SRCS proto/server_configure.proto)
 add_custom_target(server_config_py_proto_init ALL COMMAND ${CMAKE_COMMAND} -E touch __init__.py)
diff --git a/python/examples/faster_rcnn_model/new_test_client.py b/python/examples/faster_rcnn_model/new_test_client.py
index 197530a6fa0d87bfe6ca7bf57fa903479fdaa86f..0c6c615f8f3dff10626256de59101c401457509f 100755
--- a/python/examples/faster_rcnn_model/new_test_client.py
+++ b/python/examples/faster_rcnn_model/new_test_client.py
@@ -22,13 +22,19 @@ preprocess = Sequential([
 ])
 
 postprocess = RCNNPostprocess("label_list.txt", "output")
-
 client = Client()
-client.load_client_config("faster_rcnn_client_conf/serving_client_conf.prototxt")
+
+client.load_client_config(
+    "faster_rcnn_client_conf/serving_client_conf.prototxt")
 client.connect(['127.0.0.1:9393'])
 
 im = preprocess(sys.argv[2])
-fetch_map = client.predict(feed={"image": im, "im_info": np.array(list(im.shape[1:]) + [1.0]),
-                                 "im_shape": np.array(list(im.shape[1:]) + [1.0])}, fetch=["multiclass_nms"])
+fetch_map = client.predict(
+    feed={
+        "image": im,
+        "im_info": np.array(list(im.shape[1:]) + [1.0]),
+        "im_shape": np.array(list(im.shape[1:]) + [1.0])
+    },
+    fetch=["multiclass_nms"])
 fetch_map["image"] = sys.argv[1]
 postprocess(fetch_map)
diff --git a/python/examples/resnet_v2_50/resnet50_debug.py b/python/examples/resnet_v2_50/resnet50_debug.py
new file mode 100644
index 0000000000000000000000000000000000000000..62cb1812c5718ae1f9e10e9e9a57d7c1ae6736b7
--- /dev/null
+++ b/python/examples/resnet_v2_50/resnet50_debug.py
@@ -0,0 +1,31 @@
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# 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 paddle_serving_app.reader import Sequential, File2Image, Resize, CenterCrop
+from paddle_serving_app.reader import RGB2BGR, Transpose, Div, Normalize
+from paddle_serving_app import Debugger
+import sys
+
+debugger = Debugger()
+debugger.load_model_config(sys.argv[1], gpu=True)
+
+seq = Sequential([
+    File2Image(), Resize(256), CenterCrop(224), RGB2BGR(), Transpose((2, 0, 1)),
+    Div(255), Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225], True)
+])
+
+image_file = "daisy.jpg"
+img = seq(image_file)
+fetch_map = debugger.predict(feed={"image": img}, fetch=["feature_map"])
+print(fetch_map["feature_map"].reshape(-1))
diff --git a/python/paddle_serving_app/__init__.py b/python/paddle_serving_app/__init__.py
index 860876030695baee15d3ace68c6af386290cfbb0..fd9260284b4103f00ca8b9cda8b99173591d23eb 100644
--- a/python/paddle_serving_app/__init__.py
+++ b/python/paddle_serving_app/__init__.py
@@ -16,3 +16,4 @@ from .reader.image_reader import ImageReader, File2Image, URL2Image, Sequential,
 from .reader.lac_reader import LACReader
 from .reader.senta_reader import SentaReader
 from .models import ServingModels
+from .local_predict import Debugger
diff --git a/python/paddle_serving_app/local_predict.py b/python/paddle_serving_app/local_predict.py
new file mode 100644
index 0000000000000000000000000000000000000000..133aa4ccf32d29538d5b7032874f2c770e55e184
--- /dev/null
+++ b/python/paddle_serving_app/local_predict.py
@@ -0,0 +1,124 @@
+# -*- coding: utf-8 -*-
+"""
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# 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 os
+import google.protobuf.text_format
+import numpy as np
+import argparse
+import paddle.fluid as fluid
+from .proto import general_model_config_pb2 as m_config
+from paddle.fluid.core import PaddleTensor
+from paddle.fluid.core import AnalysisConfig
+from paddle.fluid.core import create_paddle_predictor
+import logging
+
+logging.basicConfig(format="%(asctime)s - %(levelname)s - %(message)s")
+logger = logging.getLogger("fluid")
+logger.setLevel(logging.INFO)
+
+
+class Debugger(object):
+    def __init__(self):
+        self.feed_names_ = []
+        self.fetch_names_ = []
+        self.feed_types_ = {}
+        self.fetch_types_ = {}
+        self.feed_shapes_ = {}
+        self.feed_names_to_idx_ = {}
+        self.fetch_names_to_idx_ = {}
+        self.fetch_names_to_type_ = {}
+
+    def load_model_config(self, model_path, gpu=False, profile=True, cpu_num=1):
+        client_config = "{}/serving_server_conf.prototxt".format(model_path)
+        model_conf = m_config.GeneralModelConfig()
+        f = open(client_config, 'r')
+        model_conf = google.protobuf.text_format.Merge(
+            str(f.read()), model_conf)
+        config = AnalysisConfig(model_path)
+
+        self.feed_names_ = [var.alias_name for var in model_conf.feed_var]
+        self.fetch_names_ = [var.alias_name for var in model_conf.fetch_var]
+        self.feed_names_to_idx_ = {}
+        self.fetch_names_to_idx_ = {}
+
+        for i, var in enumerate(model_conf.feed_var):
+            self.feed_names_to_idx_[var.alias_name] = i
+            self.feed_types_[var.alias_name] = var.feed_type
+            self.feed_shapes_[var.alias_name] = var.shape
+
+        for i, var in enumerate(model_conf.fetch_var):
+            self.fetch_names_to_idx_[var.alias_name] = i
+            self.fetch_names_to_type_[var.alias_name] = var.fetch_type
+
+        if not gpu:
+            config.disable_gpu()
+        else:
+            config.enable_use_gpu(100, 0)
+        if profile:
+            config.enable_profile()
+        config.set_cpu_math_library_num_threads(cpu_num)
+
+        self.predictor = create_paddle_predictor(config)
+
+    def predict(self, feed=None, fetch=None):
+        if feed is None or fetch is None:
+            raise ValueError("You should specify feed and fetch for prediction")
+        fetch_list = []
+        if isinstance(fetch, str):
+            fetch_list = [fetch]
+        elif isinstance(fetch, list):
+            fetch_list = fetch
+        else:
+            raise ValueError("Fetch only accepts string and list of string")
+
+        feed_batch = []
+        if isinstance(feed, dict):
+            feed_batch.append(feed)
+        elif isinstance(feed, list):
+            feed_batch = feed
+        else:
+            raise ValueError("Feed only accepts dict and list of dict")
+
+        int_slot_batch = []
+        float_slot_batch = []
+        int_feed_names = []
+        float_feed_names = []
+        int_shape = []
+        float_shape = []
+        fetch_names = []
+        counter = 0
+        batch_size = len(feed_batch)
+
+        for key in fetch_list:
+            if key in self.fetch_names_:
+                fetch_names.append(key)
+
+        if len(fetch_names) == 0:
+            raise ValueError(
+                "Fetch names should not be empty or out of saved fetch list.")
+            return {}
+
+        inputs = []
+        for name in self.feed_names_:
+            inputs.append(PaddleTensor(feed[name][np.newaxis, :]))
+
+        outputs = self.predictor.run(inputs)
+        fetch_map = {}
+        for name in fetch:
+            fetch_map[name] = outputs[self.fetch_names_to_idx_[
+                name]].as_ndarray()
+        return fetch_map
diff --git a/python/setup.py.app.in b/python/setup.py.app.in
index d981caa7c34ff1f84d4cdee0e64a2b03e47b7b66..77099e667e880f3f62ab4cde9d5ae3b6295d1b90 100644
--- a/python/setup.py.app.in
+++ b/python/setup.py.app.in
@@ -42,10 +42,11 @@ if '${PACK}' == 'ON':
 
 
 REQUIRED_PACKAGES = [
-    'six >= 1.10.0', 'sentencepiece'
+    'six >= 1.10.0', 'sentencepiece', 'opencv-python', 'pillow'
 ]
 
 packages=['paddle_serving_app',
+	  'paddle_serving_app.proto',
           'paddle_serving_app.reader',
 	  'paddle_serving_app.utils',
 	  'paddle_serving_app.models',
@@ -54,6 +55,8 @@ packages=['paddle_serving_app',
 package_data={}
 package_dir={'paddle_serving_app':
              '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving_app',
+	     'paddle_serving_app.proto':
+	     '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving_app/proto',
              'paddle_serving_app.reader':
              '${PADDLE_SERVING_BINARY_DIR}/python/paddle_serving_app/reader',
 	     'paddle_serving_app.utils':