diff --git a/tools/har-to-old.py b/tools/har-to-old.py
new file mode 100755
index 0000000000000000000000000000000000000000..fa2061641efc7513affb25adf2cd673735558a9e
--- /dev/null
+++ b/tools/har-to-old.py
@@ -0,0 +1,87 @@
+#!/usr/bin/env python3
+
+from argparse import ArgumentParser, FileType
+from base64 import b64decode
+import json
+import mimetypes
+from pathlib import Path
+from urllib.parse import urlparse
+
+
+def write_request(entry, fd):
+    entry = entry['request']
+
+    # we should put the path, but since requests does not output the Host header
+    # we would not know what was the host
+    fd.write(f"{entry['method']} {entry['url']} {entry['httpVersion']}\n".encode())
+
+    for header in entry['headers']:
+        fd.write(f"{header['name']}: {header['value']}\n".encode())
+
+    fd.write(b'\n')
+
+    if 'postData' in entry:
+        if entry['postData'].get('x-binary'):
+            # non-standard key emitted by weboob
+            body = entry['postData']['text'].encode('latin-1')
+        else:
+            body = entry['postData']['text'].encode()
+        fd.write(body)
+
+
+def write_response(entry, fd):
+    entry = entry['response']
+    fd.write(f"{entry['httpVersion']} {entry['status']} {entry['statusText']}\n")
+    for header in entry['headers']:
+        fd.write(f"{header['name']}: {header['value']}\n")
+
+
+def write_body(entry, fd):
+    entry = entry['response']
+    if entry['content'].get('encoding') == 'base64':
+        data = b64decode(entry['content']['text'])
+    else:
+        data = entry['content']['text'].encode('utf-8')
+    fd.write(data)
+
+
+def guess_extension(entry):
+    headers = entry['response']['headers']
+    ctype = next((header['value'] for header in headers if header['name'].lower() == 'content-type'), '')
+    # due to http://bugs.python.org/issue1043134
+    if ctype == 'text/plain':
+        ext = '.txt'
+    else:
+        # try to get an extension (and avoid adding 'None')
+        ext = mimetypes.guess_extension(ctype, False) or ''
+    return ext
+
+
+def main():
+    def extract(n, prefix):
+        entry = data['log']['entries'][n]
+
+        ext = guess_extension(entry)
+        name = Path(urlparse(entry['request']['url']).path).stem
+        prefix = f'{prefix}/{n + 1:03d}-{entry["response"]["status"]}{name and f"-{name}"}{ext}'
+
+        with open(f'{prefix}-request.txt', 'wb') as fd:
+            write_request(entry, fd)
+        with open(f'{prefix}.response.txt', 'w') as fd:
+            write_response(entry, fd)
+        with open(prefix, 'wb') as fd:
+            write_body(entry, fd)
+
+    parser = ArgumentParser()
+    parser.add_argument('file', type=FileType('r'), help='HAR file to extract')
+    parser.add_argument('prefix')
+    args = parser.parse_args()
+
+    data = json.load(args.file)
+    for n in range(len(data['log']['entries'])):
+        print('extracting request', n)
+        extract(n, args.prefix)
+
+
+if __name__ == '__main__':
+    main()