diff --git a/lookout/style/format/__main__.py b/lookout/style/format/__main__.py
index ef852a277..7d42be4f7 100644
--- a/lookout/style/format/__main__.py
+++ b/lookout/style/format/__main__.py
@@ -8,7 +8,7 @@
 
 def create_parser():
     parser = argparse.ArgumentParser(formatter_class=ArgumentDefaultsHelpFormatterNoNone)
-    subparsers = parser.add_subparsers(help="Commands", dest="command")
+    subparsers = parser.add_subparsers(help="Commands")
 
     def add_parser(name, help):
         return subparsers.add_parser(
@@ -17,7 +17,7 @@ def add_parser(name, help):
     # Evaluation
     eval_parser = add_parser("eval", "Evaluate trained model on given dataset.")
     eval_parser.set_defaults(handler=quality_report)
-    eval_parser.add_argument("-i", "--input", required=True, type=str,
+    eval_parser.add_argument("-i", "--input-pattern", required=True, type=str,
                              help="Path to folder with source code - "
                                   "should be in a format compatible with glob (ends with**/* "
                                   "and surrounded by quotes. Ex: `path/**/*`).")
@@ -33,8 +33,8 @@ def add_parser(name, help):
     # Visualization
     vis_parser = add_parser("vis", "Visualize mispredictions of the model on the given file.")
     vis_parser.set_defaults(handler=visualize)
-    vis_parser.add_argument("-i", "--input", required=True, help="Path to folder with source "
-                                                                 "code.")
+    vis_parser.add_argument("-i", "--input-filename", required=True,
+                            help="Path to file to analyze.")
     vis_parser.add_argument("--bblfsh", default="0.0.0.0:9432",
                             help="Babelfish server's address.")
     vis_parser.add_argument("-l", "--language", default="javascript",
@@ -48,12 +48,13 @@ def main():
     args = parser.parse_args()
     try:
         handler = args.handler
+        delattr(args, "handler")
     except AttributeError:
         def print_usage(_):
             parser.print_usage()
 
         handler = print_usage
-    return handler(args)
+    return handler(**vars(args))
 
 
 if __name__ == "__main__":
diff --git a/lookout/style/format/quality_report.py b/lookout/style/format/quality_report.py
index 49fa92c1a..5bd81e99d 100644
--- a/lookout/style/format/quality_report.py
+++ b/lookout/style/format/quality_report.py
@@ -2,7 +2,7 @@
 import glob
 import os
 
-import bblfsh
+from bblfsh import BblfshClient
 from bblfsh.client import NonUTF8ContentException
 import numpy
 from sklearn.metrics import classification_report,confusion_matrix
@@ -36,16 +36,17 @@ def prepare_files(folder, client, language):
     return files
 
 
-def quality_report(args):
-    client = bblfsh.BblfshClient(args.bblfsh)
-    files = prepare_files(args.input, client, args.language)
+def quality_report(input_pattern: str, bblfsh: str, language: str, n_files: int,
+                   model: str) -> None:
+    client = BblfshClient(bblfsh)
+    files = prepare_files(input_pattern, client, language)
     print("Number of files: %s" % (len(files)))
 
-    fe = FeatureExtractor(language=args.language)
+    fe = FeatureExtractor(language=language)
     X, y, nodes = fe.extract_features(files)
 
-    analyzer = FormatModel().load(args.model)
-    rules = analyzer._rules_by_lang[args.language]
+    analyzer = FormatModel().load(model)
+    rules = analyzer._rules_by_lang[language]
     y_pred = rules.predict(X)
 
     target_names = [CLASSES[cls_ind] for cls_ind in numpy.unique(y)]
@@ -60,7 +61,7 @@ def quality_report(args):
     file_stat = Counter(file_mispred)
 
     to_show = file_stat.most_common()
-    if args.n_files > 0:
-        to_show = to_show[:args.n_files]
+    if n_files > 0:
+        to_show = to_show[:n_files]
 
     print("Files with most errors:\n" + "\n".join(map(str, to_show)))
diff --git a/lookout/style/format/visualization.py b/lookout/style/format/visualization.py
index dd7b3e5b5..091b7a39d 100644
--- a/lookout/style/format/visualization.py
+++ b/lookout/style/format/visualization.py
@@ -1,7 +1,7 @@
 from collections import namedtuple
 import os
 
-import bblfsh
+from bblfsh import BblfshClient
 
 from lookout.core.api.service_data_pb2 import File
 from lookout.style.format.features import FeatureExtractor, CLASSES
@@ -28,15 +28,15 @@ def prepare_file(filename, client, language):
     return File(content=content, uast=res.uast, path=filename)
 
 
-def visualize(args):
-    client = bblfsh.BblfshClient(args.bblfsh)
-    file = prepare_file(args.input, client, args.language)
+def visualize(input_filename: str, bblfsh: str, language: str, model: str) -> None:
+    client = BblfshClient(bblfsh)
+    file = prepare_file(input_filename, client, language)
 
-    fe = FeatureExtractor(language=args.language)
+    fe = FeatureExtractor(language=language)
     X, y, nodes = fe.extract_features([file])
 
-    analyzer = FormatModel().load(args.model)
-    rules = analyzer._rules_by_lang[args.language]
+    analyzer = FormatModel().load(model)
+    rules = analyzer._rules_by_lang[language]
     y_pred = rules.predict(X)
 
     mispred = []