diff --git a/.envrc b/.envrc
new file mode 100644
index 0000000..298b682
--- /dev/null
+++ b/.envrc
@@ -0,0 +1,5 @@
+export AWS_ACCESS_KEY_ID="FANCY-ROOT-KEY"
+export AWS_SECRET_ACCESS_KEY="SECRET"
+
+# Optional: Run a command automatically when .envrc loads
+# aws s3 ls
diff --git a/README.md b/README.md
index 23d5132..d38659b 100644
--- a/README.md
+++ b/README.md
@@ -12,4 +12,20 @@ Simply download suitable sample script for your use-case and run it close to the
 
 Currently list of available sample scripts is rather short, but we are working on extending it. Star to follow on our updates.
 
-For additional informaition, please refer to our documentation: https://docs.ultihash.io/
\ No newline at end of file
+For additional informaition, please refer to our documentation: https://docs.ultihash.io/
+
+Upload a folder:
+
+```bash
+./boto3/multithread_upload/uh_upload.py --url http://localhost:8080 -B xxx -e <source_folder>
+```
+
+Download the bucket:
+
+```bash
+./boto3/multithread_download/uh_download.py xxx --url http://localhost:8080 -C <target_folder> --delete-path
+```
+
+You'd better use ramdisk folder as a target folder for download, as it will be much faster and perform stabler.
+
+Adding --delete-path option will remove the target folder before download, so you're download doesn't do any hidden remove operation which may slow down the process.
diff --git a/boto3/multithread_download/uh_download.py b/boto3/multithread_download/uh_download.py
index b73c2b0..425b8ee 100755
--- a/boto3/multithread_download/uh_download.py
+++ b/boto3/multithread_download/uh_download.py
@@ -8,6 +8,9 @@
 import sys
 import time
 import tqdm
+import os
+
+import shutil
 
 def parse_args():
     parser = argparse.ArgumentParser(
@@ -31,6 +34,9 @@ def parse_args():
     parser.add_argument('--no-store', help='do not store downloaded data to file system, only used for benchmarking',
                         action='store_false', dest='store')
 
+    parser.add_argument('--delete-path', help='recursively delete target path before download',
+                        action='store_true', dest='delete_path')
+
     return parser.parse_args()
 
 
@@ -59,13 +65,22 @@ def cb(count):
 
 
         response = self.s3.get_object(Bucket=bucket, Key=key)
-        body = response["Body"].read()
-        cb(len(body))
-
+        stream = response["Body"]
         if self.config.store:
             local_path.parent.mkdir(parents=True, exist_ok=True)
             with open(local_path, "wb+") as f:
-                f.write(body)
+                while True:
+                    chunk = stream.read(1048576)
+                    if not chunk:
+                        break
+                    f.write(chunk)
+                    cb(len(chunk))
+        else:
+            while True:
+                chunk = stream.read(1048576)
+                if not chunk:
+                    break
+                cb(len(chunk))
 
     def list_objects(self, bucket):
         paginator = self.s3.get_paginator('list_objects_v2')
@@ -117,4 +132,12 @@ def download(config):
 
 if __name__ == "__main__":
     config = parse_args()
+
+    def delete_path(config):
+        if config.path.exists():
+            shutil.rmtree(config.path)
+            os.sync()
+
+    if config.delete_path:
+        delete_path(config)
     download(config)
diff --git a/boto3/multithread_upload/uh_upload.py b/boto3/multithread_upload/uh_upload.py
index 4cfd7ba..ad670d6 100755
--- a/boto3/multithread_upload/uh_upload.py
+++ b/boto3/multithread_upload/uh_upload.py
@@ -44,6 +44,9 @@ def parse_args():
     parser.add_argument('--generate', help='generate and upload random data of the specified size in GiB',
                         action='store', dest='generate', type=int)
 
+    parser.add_argument('-e', '--check-etag', help='check ETag after upload (single part only)',
+        action='store_true', dest='check_etag')
+
     args, unknown = parser.parse_known_args()
 
     if args.generate:
@@ -104,7 +107,9 @@ def cb(count):
                 self.progress.update(count)
             else:
                 self.count_buffer += count
-        self.s3.upload_file(file_path, Bucket=bucket, Key=str(file_path.relative_to(base_path)), Callback=cb, Config=self.transfer_config)
+        key = str(file_path.relative_to(base_path))
+        self.s3.upload_file(file_path, Bucket=bucket, Key=key, Callback=cb, Config=self.transfer_config)
+        return key
 
     def upload_random(self, bucket, key, size):
         def cb(count):
@@ -180,23 +185,38 @@ def upload (config):
                 pass
 
             if base_path.is_file():
-                results += [(base_path, up.push(bucket, base_path, pathlib.Path(base_path).parent))]
-                size_total += base_path.stat().st_size
-                continue
+                  key = str(base_path.relative_to(pathlib.Path(base_path).parent))
+                  future = up.push(bucket, base_path, pathlib.Path(base_path).parent)
+                  results += [(base_path, future, bucket, key)]
+                  size_total += base_path.stat().st_size
+                  continue
 
             for (root, dirs, files) in os.walk(base_path):
-                for file in files:
-                    file_path = pathlib.Path(root) / file
-                    size_total += file_path.stat().st_size
-                    results += [(file_path, up.push(bucket, file_path, base_path))]
+                  for file in files:
+                      file_path = pathlib.Path(root) / file
+                      key = str(file_path.relative_to(base_path))
+                      future = up.push(bucket, file_path, base_path)
+                      results += [(file_path, future, bucket, key)]
+                      size_total += file_path.stat().st_size
 
     up.set_total(size_total)
 
+    import hashlib
+    def calc_md5(path):
+        h = hashlib.md5()
+        with open(path, 'rb') as f:
+            for chunk in iter(lambda: f.read(8192), b''):
+                h.update(chunk)
+        return h.hexdigest()
+
+    # Wait for all uploads to finish
     for job in results:
-        try:
-            job[1].result()
-        except Exception as e:
-            print(f"Error uploading {job[0]}: {str(e)}", file=sys.stderr)
+          # Wait for each upload to finish
+          file_path, future, bucket, key = job
+          try:
+              future.result()
+          except Exception as e:
+              print(f"Error uploading {file_path}: {str(e)}", file=sys.stderr)
 
     end = time.monotonic()
     seconds = end - start
@@ -205,6 +225,30 @@ def upload (config):
     up.stop()
     print(f"average upload speed: {mb/seconds} MB/s")
 
+    # ETag check after timing
+    etag_fail = False
+    for job in results:
+        file_path, _, bucket, key = job
+        try:
+            if config.check_etag and pathlib.Path(file_path).is_file():
+                resp = up.s3.head_object(Bucket=bucket, Key=key)
+                etag = resp['ETag'].strip('"')
+                md5 = calc_md5(file_path)
+                if '-' in etag:
+                    print(f"[ETag] {file_path}: multipart upload detected, cannot compare MD5.")
+                elif etag == md5:
+                    print(f"[ETag] {file_path}: OK (etag matches local MD5)")
+                else:
+                    print(f"\033[31m[ETag] {file_path}: FAIL (etag {etag} != local MD5 {md5})\033[0m")
+                    etag_fail = True
+            else:
+                pass
+        except Exception as e:
+            print(f"Error uploading {file_path}: {str(e)}", file=sys.stderr)
+            etag_fail = True
+
+    if config.check_etag and etag_fail:
+        sys.exit(1)
     return float(mb)/seconds
 
 if __name__ == "__main__":