diff --git a/.flake8 b/.flake8
index a1288dba5..d0c265bc1 100644
--- a/.flake8
+++ b/.flake8
@@ -15,7 +15,11 @@
 
 [flake8]
 ignore = F401, F811, F821, E127, E128, E301, E302, E305, E501, E701, E704, E999, B303
+# We are checking with Python 3 but many of the stubs are Python 2 stubs.
+# A nice future improvement would be to provide separate .flake8
+# configurations for Python 2 and Python 3 files.
+builtins = StandardError,apply,basestring,buffer,cmp,coerce,execfile,file,intern,long,raw_input,reduce,reload,unichr,unicode,xrange
 
 # Errors that we need to fix before enabling flake8 by default:
-#   936 F821 undefined name - FIXME: re-enable and add missing imports
+#   208 F821 undefined name - FIXME: re-enable and add missing imports
 #     6 E999 invalid syntax - FIXME: re-enable after flake8 is running from Python 3.6
diff --git a/.travis.yml b/.travis.yml
index 1db0ec621..9e6a51463 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -4,7 +4,7 @@ language: python
 matrix:
   include:
     - python: "3.5"
-      env: TEST_CMD="./tests/flake8_test.py"
+      env: TEST_CMD="flake8"
     - python: "3.5"
       env: TEST_CMD="./tests/mypy_test.py"
     - python: "2.7"
@@ -12,7 +12,7 @@ matrix:
 
 install:
   # pytype needs py-2.7, mypy needs py-3.2+. Additional logic in runtests.py
-  - if [[ $TRAVIS_PYTHON_VERSION == '3.5' ]]; then pip install -U git+git://github.com/python/mypy git+git://github.com/dropbox/typed_ast flake8-bugbear; fi
+  - if [[ $TRAVIS_PYTHON_VERSION == '3.5' ]]; then pip install -U git+git://github.com/python/mypy git+git://github.com/dropbox/typed_ast flake8==3.2.1 flake8-bugbear flake8-pyi; fi
   - if [[ $TRAVIS_PYTHON_VERSION == '2.7' ]]; then pip install -U git+git://github.com/google/pytype; fi
 
 script:
diff --git a/README.md b/README.md
index adc239748..4fbe1a495 100644
--- a/README.md
+++ b/README.md
@@ -116,7 +116,7 @@ mypy tests and flake8 tests by invoking:
 ```
 (.venv3)$ python tests/mypy_test.py
 ...
-(.venv3)$ python tests/flake8_test.py
+(.venv3)$ flake8
 ...
 ```
 To run the pytype tests, you need a separate virtual environment with
@@ -146,7 +146,7 @@ running mypy --python-version 3.2 --strict-optional # with 453 files
 running mypy --python-version 2.7 --strict-optional # with 502 files
 Running pytype tests...
 Ran pytype with 244 pyis, got 0 errors.
-Running flake8 on 886 .pyi files...
+Running flake8...
 flake8 run clean.
 (.venv3)$
 ```
diff --git a/requirements-tests-py3.txt b/requirements-tests-py3.txt
index 28aa0670c..66d4a2077 100644
--- a/requirements-tests-py3.txt
+++ b/requirements-tests-py3.txt
@@ -1,4 +1,5 @@
 mypy-lang>=0.4.6
 typed-ast>=0.6.1
-flake8>=3.2.1
+flake8==3.2.1
 flake8-bugbear>=16.12.2
+flake8-pyi>=16.12.0
diff --git a/runtests.sh b/runtests.sh
index 4646841d2..164506027 100755
--- a/runtests.sh
+++ b/runtests.sh
@@ -2,4 +2,5 @@
 
 ./tests/mypy_test.py
 ./tests/pytype_test.py
-./tests/flake8_test.py
+echo "Running flake8..."
+flake8 && echo "flake8 run clean."
diff --git a/tests/flake8_test.py b/tests/flake8_test.py
deleted file mode 100755
index 2f00328fc..000000000
--- a/tests/flake8_test.py
+++ /dev/null
@@ -1,47 +0,0 @@
-#!/usr/bin/env python3
-import pathlib
-import subprocess
-import sys
-
-PY2_ONLY_KEYWORDS = [
-    'StandardError',
-    'apply',
-    'basestring',
-    'buffer',
-    'cmp',
-    'coerce',
-    'execfile',
-    'file',
-    'intern',
-    'long',
-    'raw_input',
-    'reduce',
-    'reload',
-    'unichr',
-    'unicode',
-    'xrange',
-]
-
-root = pathlib.Path(__file__).parent.parent
-paths = list(sorted(str(p) for p in root.glob('**/*.pyi')))
-window = 0
-size = 100
-returncode = 0
-
-print('Running flake8 on {} .pyi files...'.format(len(paths)))
-
-while True:
-    chunk = paths[window:window + size]
-    if not chunk:
-        break
-
-    proc = subprocess.run(
-        ['flake8', '--builtins=' + ','.join(PY2_ONLY_KEYWORDS)] + chunk,
-    )
-    if proc.returncode:
-        print('flake8 run failed!')
-        sys.exit(1)
-
-    window += size
-
-print('flake8 run clean.')
diff --git a/tests/mypy_test.py b/tests/mypy_test.py
index 3498059c4..d474009e4 100755
--- a/tests/mypy_test.py
+++ b/tests/mypy_test.py
@@ -126,7 +126,7 @@ def main():
             flags.append('--strict-optional')
             if (major, minor) >= (3, 6):
                 flags.append('--fast-parser')
-            ##flags.append('--warn-unused-ignores')  # Fast parser and regular parser disagree.
+            # flags.append('--warn-unused-ignores')  # Fast parser and regular parser disagree.
             sys.argv = ['mypy'] + flags + files
             if args.verbose:
                 print("running", ' '.join(sys.argv))