diff --git a/runner.py b/runner.py
new file mode 100644
index 0000000..21e93f6
--- /dev/null
+++ b/runner.py
@@ -0,0 +1,67 @@
+#!/usr/bin/python3
+import argparse
+import os
+import subprocess
+
+DESCRIPTION = '''sqlite-parquet-vtable test runner
+Run a query against several data files with varying encodings,
+and verify that the output matches an expected value
+'''
+
+COMMON_QUERY_LINES = [
+        '.echo off',
+        '.load ./libparquet',
+]
+
+
+def read(filename):
+    '''Take a filename, read it into a variable -- critically, closing it after we're done'''
+    with open(filename) as file:
+        return file.read()
+
+
+def dispatch(query_file, results_file, expected_exit_code, datasets):
+    '''Run a query against assorted formats of the datasets, verifying the result each time'''
+    query = read(query_file)
+    expected_results = read(results_file)
+    for dataset in datasets:
+        for file in os.listdir(dataset):
+            if not file.endswith('.parquet'):
+                print(f'Ignoring {file} -- does not end in .parquet')
+                continue
+            vtable_statement = f'CREATE VIRTUAL TABLE dataset USING parquet(\'{file}\');'
+            # Append test-specified query to common lines, insert \n between lines
+            full_query = '\n'.join(COMMON_QUERY_LINES+[vtable_statement, query])
+            proc = subprocess.run(
+                'sqlite3',
+                stdout=subprocess.PIPE,
+                input=full_query,
+                encoding='UTF-8',
+                check=False
+            )
+            assert proc.returncode == expected_exit_code
+            assert proc.stdout == expected_results
+
+
+if __name__ == '__main__':
+    PARSER = argparse.ArgumentParser(description=DESCRIPTION)
+    PARSER.add_argument('query', metavar='query', help='the .sql file containing the query')
+    PARSER.add_argument('results', help='the file containing .output format expected results')
+    PARSER.add_argument('exit_code', help='the exit code you expect to recieve')
+    PARSER.add_argument('dataset', nargs='+', help='a dataset to run the query against')
+
+    ARGS = PARSER.parse_args()
+
+    # Verify that each query and result file exist
+    assert os.path.isfile(ARGS.query)
+    assert os.path.isfile(ARGS.results)
+    # Verify that each dataset argument is a folder (ideally with .parquet files inside)
+    for dataset_dir in ARGS.dataset:
+        assert os.path.isdir(dataset_dir)
+
+    dispatch(
+        query_file=ARGS.query,
+        results_file=ARGS.results,
+        expected_exit_code=ARGS.exit_code,
+        datasets=ARGS.dataset
+    )