diff --git a/.circleci/config.yml b/.circleci/config.yml
index f32701076e..504ae22f87 100644
--- a/.circleci/config.yml
+++ b/.circleci/config.yml
@@ -159,13 +159,6 @@ step-nodejs-headers-build: &step-nodejs-headers-build
       cd src
       ninja -C out/Default third_party/electron_node:headers
 
-step-native-tests-build: &step-native-tests-build
-  run:
-    name: Native tests build
-    command: |
-      cd src
-      ninja -C out/Default electron:electron_tests
-
 step-persist-data-for-tests: &step-persist-data-for-tests
   persist_to_workspace:
     root: .
@@ -346,9 +339,34 @@ steps-native-tests: &steps-native-tests
     - *step-setup-env-for-build
 
     - *step-electron-gn-gen
-    - *step-native-tests-build
-
-    # TODO(alexeykuzmin): Run the tests. It can be extremely parallelized!
+    - run:
+        name: List native tests
+        command: |
+          cd src
+          python electron/script/native-tests.py list \
+            --config electron/script/tests.yml > testslist.txt
+    # TODO(alexeykuzmin): Build only a subset of all tests.
+    - run:
+        name: Build native tests
+        command: |
+          cd src
+          ninja -C out/Default electron:electron_tests
+    # TODO(alexeykuzmin): Run only a subset of all tests.
+    # --binary $(circleci tests split testslist.txt)
+    - run:
+        name: Run native tests
+        command: |
+          cd src
+          mkdir tests_results
+          python electron/script/native-tests.py run \
+            --config electron/script/tests.yml \
+            --tests-dir out/Default \
+            --output-dir tests_results
+    - store_artifacts:
+        path: src/tests_results
+        destination: results
+    - store_test_results:
+        path: src/tests_results
 
 steps-verify-ffmpeg: &steps-verify-ffmpeg
   steps:
@@ -590,11 +608,13 @@ jobs:
     <<: *steps-build-mac
 
   # Layer 3: Tests.
-  linux-x64-native-tests-fyi:
+  linux-x64-native-tests:
     <<: *machine-linux-2xlarge
     environment:
       <<: *env-testing-build
     <<: *steps-native-tests
+    # TODO(alexeykuzmin): Use parallelism.
+    # parallelism: 4  # https://xkcd.com/221/
 
   linux-x64-testing-tests:
     <<: *machine-linux-medium
@@ -642,10 +662,6 @@ workflows:
       - linux-arm-checkout
       - linux-arm64-checkout
 
-      - linux-x64-native-tests-fyi:
-          requires:
-            - linux-checkout
-
       - linux-x64-debug:
           requires:
             - linux-checkout
@@ -730,6 +746,10 @@ workflows:
       - linux-arm-checkout
       - linux-arm64-checkout
 
+      - linux-x64-native-tests:
+          requires:
+            - linux-checkout
+
       - linux-x64-release:
           requires:
             - linux-checkout
diff --git a/script/native-tests.py b/script/native-tests.py
new file mode 100755
index 0000000000..d0be0db1b2
--- /dev/null
+++ b/script/native-tests.py
@@ -0,0 +1,230 @@
+#!/usr/bin/env python
+
+import argparse
+import os
+import subprocess
+import sys
+
+SOURCE_ROOT = os.path.abspath(os.path.dirname(os.path.dirname(__file__)))
+VENDOR_DIR = os.path.join(SOURCE_ROOT, 'vendor')
+PYYAML_LIB_DIR = os.path.join(VENDOR_DIR, 'pyyaml', 'lib')
+sys.path.append(PYYAML_LIB_DIR)
+import yaml  #pylint: disable=wrong-import-position,wrong-import-order
+
+
+class Command:
+  LIST = 'list'
+  RUN = 'run'
+
+def parse_args():
+  parser = argparse.ArgumentParser(description='Run Google Test binaries')
+
+  parser.add_argument('command',
+                      choices=[Command.LIST, Command.RUN],
+                      help='command to execute')
+
+  parser.add_argument('-b', '--binary', nargs='*', required=False,
+                      help='names of binaries to run')
+  parser.add_argument('-c', '--config', required=True,
+                      help='path to a tests config')
+  parser.add_argument('-t', '--tests-dir', required=False,
+                      help='path to a directory with binaries to run')
+  parser.add_argument('-o', '--output-dir', required=False,
+                      help='path to a folder to save tests results')
+
+  args = parser.parse_args()
+
+  # Additional checks.
+  if args.command == Command.RUN and args.tests_dir is None:
+    parser.error("specify a path to a dir with test binaries via --tests-dir")
+
+  # Absolutize and check paths.
+  # 'config' must exist and be a file.
+  args.config = os.path.abspath(args.config)
+  if not os.path.isfile(args.config):
+    parser.error("file '{}' doesn't exist".format(args.config))
+
+  # 'tests_dir' must exist and be a directory.
+  if args.tests_dir is not None:
+    args.tests_dir = os.path.abspath(args.tests_dir)
+    if not os.path.isdir(args.tests_dir):
+      parser.error("directory '{}' doesn't exist".format(args.tests_dir))
+
+  # 'output_dir' must exist and be a directory.
+  if args.output_dir is not None:
+    args.output_dir = os.path.abspath(args.output_dir)
+    if not os.path.isdir(args.output_dir):
+      parser.error("directory '{}' doesn't exist".format(args.output_dir))
+
+  return args
+
+
+def main():
+  args = parse_args()
+  tests_list = TestsList(args.config, args.tests_dir)
+
+  if args.command == Command.LIST:
+    all_binaries_names = tests_list.get_names()
+    print '\n'.join(all_binaries_names)
+    return 0
+
+  if args.command == Command.RUN:
+    if args.binary is not None:
+      return tests_list.run(args.binary, args.output_dir)
+    else:
+      return tests_list.run_all(args.output_dir)
+
+  raise Exception("unexpected command '{}'".format(args.command))
+
+
+class TestsList():
+  def __init__(self, config_path, tests_dir):
+    self.config_path = config_path
+    self.tests_dir = tests_dir
+
+    # A dict with binary names (e.g. 'base_unittests') as keys
+    # and various test data as values of dict type.
+    self.tests = self.__get_tests_list(config_path)
+
+  def __len__(self):
+    return len(self.tests)
+
+  def get_names(self):
+    return self.tests.keys()
+
+  def run(self, binaries, output_dir=None):
+    # Don't run anything twice.
+    binaries = set(binaries)
+
+    # First check that all names are present in the config.
+    if any([binary_name not in self.tests for binary_name in binaries]):
+      raise Exception("binary '{0}' not found in config '{1}'".format(
+          binary_name, self.config_path))
+
+    # TODO(alexeykuzmin): Respect the "platform" setting.
+
+    suite_returncode = sum(
+        [self.__run(binary, output_dir) for binary in binaries])
+    return suite_returncode
+
+  def run_only(self, binary_name, output_dir=None):
+    return self.run([binary_name], output_dir)
+
+  def run_all(self, output_dir=None):
+    return self.run(self.get_names(), output_dir)
+
+  def __get_tests_list(self, config_path):
+    tests_list = {}
+    config_data = TestsList.__get_config_data(config_path)
+
+    for data_item in config_data['tests']:
+      (binary_name, test_data) = TestsList.__get_test_data(data_item)
+      tests_list[binary_name] = test_data
+
+    return tests_list
+
+  @staticmethod
+  def __get_config_data(config_path):
+    with open(config_path, 'r') as stream:
+      return yaml.load(stream)
+
+  @staticmethod
+  def __expand_shorthand(value):
+    """ Treat a string as {'string_value': None}."""
+    if isinstance(value, dict):
+      return value
+
+    if isinstance(value, basestring):
+      return {value: None}
+
+    assert False, "unexpected shorthand type: {}".format(type(value))
+
+  @staticmethod
+  def __make_a_list(value):
+    """Make a list if not already a list."""
+    if isinstance(value, list):
+        return value
+    return [value]
+
+  @staticmethod
+  def __get_test_data(data_item):
+    data_item = TestsList.__expand_shorthand(data_item)
+
+    binary_name = data_item.keys()[0]
+    test_data = {
+      'excluded_tests': None,
+      'platforms': None  # None means all? Weird.
+    }
+
+    configs = data_item[binary_name]
+    if configs is not None:
+      # List of excluded tests.
+      if 'to_fix' in configs:
+        test_data['excluded_tests'] = configs['to_fix']
+
+      # TODO(alexeykuzmin): Also add to 'excluded_tests'
+      # those that should be permanently disabled.
+
+      # List of platforms to run the tests on.
+      if 'platform' in configs:
+          test_data['platforms'] = TestsList.__make_a_list(configs['platform'])
+
+    return (binary_name, test_data)
+
+  def __run(self, binary_name, output_dir):
+    binary_path = os.path.join(self.tests_dir, binary_name)
+    test_binary = TestBinary(binary_path)
+
+    test_data = self.tests[binary_name]
+    excluded_tests = test_data['excluded_tests']
+
+    output_file_path = TestsList.__get_output_path(binary_name, output_dir)
+
+    return test_binary.run(excluded_tests=excluded_tests,
+                           output_file_path=output_file_path)
+
+  @staticmethod
+  def __get_output_path(binary_name, output_dir=None):
+    if output_dir is None:
+      return None
+
+    return os.path.join(output_dir, "results_{}.xml".format(binary_name))
+
+
+class TestBinary():
+  def __init__(self, binary_path):
+    self.binary_path = binary_path
+
+    # Is only used when writing to a file.
+    self.output_format = 'xml'
+
+  def run(self, excluded_tests=None, output_file_path=None):
+    gtest_filter = ""
+    if excluded_tests is not None and len(excluded_tests) > 0:
+      excluded_tests_string = TestBinary.__format_excluded_tests(
+          excluded_tests)
+      gtest_filter = "--gtest_filter={}".format(excluded_tests_string)
+
+    gtest_output = ""
+    if output_file_path is not None:
+      gtest_output = "--gtest_output={0}:{1}".format(self.output_format,
+                                                     output_file_path)
+
+    args = [self.binary_path, gtest_filter, gtest_output]
+
+    # Suppress stdout if we're writing results to a file.
+    stdout = None
+    if output_file_path is not None:
+      devnull = open(os.devnull, 'w')
+      stdout = devnull
+
+    returncode = subprocess.call(args, stdout=stdout)
+    return returncode
+
+  @staticmethod
+  def __format_excluded_tests(excluded_tests):
+    return "-" + ":".join(excluded_tests)
+
+
+if __name__ == '__main__':
+  sys.exit(main())
diff --git a/script/tests.yml b/script/tests.yml
new file mode 100644
index 0000000000..f617ccaaee
--- /dev/null
+++ b/script/tests.yml
@@ -0,0 +1,47 @@
+tests:
+  - base_unittests:
+      to_fix:
+        - WeakPtrDeathTest.NonOwnerThreadDeletesObjectAfterReference
+        - WeakPtrDeathTest.NonOwnerThreadDeletesWeakPtrAfterReference
+        - WeakPtrDeathTest.NonOwnerThreadDereferencesWeakPtrAfterReference
+        - WeakPtrDeathTest.NonOwnerThreadReferencesObjectAfterDeletion
+        - WeakPtrDeathTest.WeakPtrCopyDoesNotChangeThreadBinding
+  - cc_unittests
+  - cc_blink_unittests
+  - content_unittests
+  - crypto_unittests
+  - device_unittests
+  - gin_unittests
+  - gpu_unittests
+  - ipc_tests
+  - media_unittests
+  - capture_unittests
+  - midi_unittests
+  - media_mojo_unittests
+  - mojo_unittests
+  - mojo_common_unittests
+  - net_unittests
+  - ppapi_unittests
+  - printing_unittests
+  - skia_unittests
+  - sql_unittests
+  - storage_unittests
+  - angle_unittests
+  - env_chromium_unittests
+  - gn_unittests
+  - ui_base_unittests
+  - compositor_unittests
+  - display_unittests
+  - events_unittests
+  - gl_unittests
+  - url_unittests
+  - url_ipc_unittests
+  - unittests
+  - content_browsertests
+  # TODO(alexeykuzmin): Do not ignore `platform` setting.
+  - disk_cache_memory_test:
+      platform: linux
+  - sandbox_linux_unittests:
+      platform: linux
+  - dbus_unittests:
+      platform: linux