diff --git a/examples/agent/README.md b/examples/agent/README.md
index 07265d9c52fa8..3e515ad1a42aa 100644
--- a/examples/agent/README.md
+++ b/examples/agent/README.md
@@ -10,7 +10,7 @@
   # Nous Hermes 2 Pro Llama 3 8B
   ./llama-server --jinja -fa --verbose \
     -hfr NousResearch/Hermes-2-Pro-Llama-3-8B-GGUF -hff Hermes-2-Pro-Llama-3-8B-Q8_0.gguf \
-    --chat-template-file tests/chat/templates/NousResearch-Hermes-2-Pro-Llama-3-8B-tool_use.jinja
+    --chat-template "$( python scripts/get_hf_chat_template.py NousResearch/Hermes-2-Pro-Llama-3-8B tool_use )"
 
   # Llama 3.1 8B
   ./llama-server --jinja -fa --verbose \
@@ -23,25 +23,25 @@
   # functionary-small-v3
   ./llama-server --jinja -fa --verbose \
     -hfr meetkai/functionary-small-v3.2-GGUF -hff functionary-small-v3.2.Q4_0.gguf \
-    --chat-template-file tests/chat/templates/meetkai-functionary-medium-v3.2.jinja
+    --chat-template "$( python scripts/get_hf_chat_template.py meetkai/functionary-medium-v3.2 )"
 
   ./llama-server --jinja -fa --verbose \
     -m ~/Downloads/functionary-small-v3.2.Q4_0.gguf \
-    --chat-template-file tests/chat/templates/meetkai-functionary-medium-v3.2.jinja
+    --chat-template "$( python scripts/get_hf_chat_template.py meetkai/functionary-medium-v3.2 )"
 
   # Llama 3.2 3B (poor adherence)
   ./llama-server --jinja -fa --verbose \
     -hfr lmstudio-community/Llama-3.2-3B-Instruct-GGUF -hff Llama-3.2-3B-Instruct-Q6_K_L.gguf \
-    --chat-template-file tests/chat/templates/meta-llama-Llama-3.2-3B-Instruct.jinja
+    --chat-template "$( python scripts/get_hf_chat_template.py meta-llama/Llama-3.2-3B-Instruct )"
 
   ./llama-server --jinja -fa --verbose \
     -m ~/Downloads/Llama-3.2-3B-Instruct-Q6_K_L.gguf \
-    --chat-template-file tests/chat/templates/meta-llama-Llama-3.2-3B-Instruct.jinja
+    --chat-template "$( python scripts/get_hf_chat_template.py meta-llama/Llama-3.2-3B-Instruct )"
 
   # Llama 3.2 1B (very poor adherence)
   ./llama-server --jinja -fa --verbose \
     -hfr lmstudio-community/Llama-3.2-1B-Instruct-GGUF -hff Llama-3.2-1B-Instruct-Q4_K_M.gguf \
-    --chat-template-file tests/chat/templates/meta-llama-Llama-3.2-3B-Instruct.jinja
+    --chat-template "$( python scripts/get_hf_chat_template.py meta-llama/Llama-3.2-3B-Instruct )"
   ```
 
 - Run the tools in [examples/agent/tools](./examples/agent/tools) inside a docker container (check http://localhost:8088/docs once running):
diff --git a/scripts/get_hf_chat_template.py b/scripts/get_hf_chat_template.py
new file mode 100644
index 0000000000000..49d0500253444
--- /dev/null
+++ b/scripts/get_hf_chat_template.py
@@ -0,0 +1,69 @@
+'''
+  Fetches the Jinja chat template of a HuggingFace model.
+  If a model 
+
+  Syntax:
+    get_hf_chat_template.py model_id [variant]
+
+  Examples:
+    python ./scripts/get_hf_chat_template.py NousResearch/Meta-Llama-3-8B-Instruct
+    python ./scripts/get_hf_chat_template.py NousResearch/Hermes-3-Llama-3.1-70B tool_use
+    python ./scripts/get_hf_chat_template.py meta-llama/Llama-3.2-3B-Instruct
+'''
+
+import json
+import re
+import sys
+
+
+def main(args):
+    if len(args) < 1:
+        raise ValueError("Please provide a model ID and an optional variant name")
+    model_id = args[0]
+    variant = None if len(args) < 2 else args[1]
+    
+    try:
+        # Use huggingface_hub library if available.
+        # Allows access to gated models if the user has access and ran `huggingface-cli login`.
+        from huggingface_hub import hf_hub_download
+        with open(hf_hub_download(repo_id=model_id, filename="tokenizer_config.json")) as f:
+            config_str = f.read()
+    except ImportError:
+        import requests
+        assert re.match(r"^[\w.-]+/[\w.-]+$", model_id), f"Invalid model ID: {model_id}"
+        response = requests.get(f"https://huggingface.co/{model_id}/resolve/main/tokenizer_config.json")
+        if response.status_code == 401:
+            raise Exception('Access to this model is gated, please request access, authenticate with `huggingface-cli login` and make sure to run `pip install huggingface_hub`')
+        response.raise_for_status()
+        config_str = response.text
+
+    try:
+        config = json.loads(config_str)
+    except json.JSONDecodeError:
+        # Fix https://huggingface.co/NousResearch/Meta-Llama-3-8B-Instruct/blob/main/tokenizer_config.json
+        # (Remove extra '}' near the end of the file)
+        config = json.loads(re.sub(r'\}([\n\s]*\}[\n\s]*\],[\n\s]*"clean_up_tokenization_spaces")', r'\1', config_str))
+
+    chat_template = config['chat_template']
+    if isinstance(chat_template, str):
+        print(chat_template, end=None)
+    else:
+        variants = {
+            ct['name']: ct['template']
+            for ct in chat_template
+        }
+        format_variants = lambda: ', '.join(f'"{v}"' for v in variants.keys())
+    
+        if variant is None:
+            if 'default' not in variants:
+                raise Exception(f'Please specify a chat template variant (one of {format_variants()})')
+            variant = 'default'
+            print(f'Note: picked "default" chat template variant (out of {format_variants()})', file=sys.stderr)
+        elif variant not in variants:
+            raise Exception(f"Variant {variant} not found in chat template (found {format_variants()})")
+        
+        print(variants[variant], end=None)
+
+
+if __name__ == '__main__':
+    main(sys.argv[1:])
diff --git a/tests/update_jinja_goldens.py b/scripts/update_jinja_goldens.py
similarity index 99%
rename from tests/update_jinja_goldens.py
rename to scripts/update_jinja_goldens.py
index 16f9c904b9452..3570c52437006 100644
--- a/tests/update_jinja_goldens.py
+++ b/scripts/update_jinja_goldens.py
@@ -10,7 +10,7 @@
   Fetches the Jinja2 templates of a few known models and use them to generate prompt goldens for a few predefined chat contexts.
 
   Examples:
-    python ./tests/update_jinja_goldens.py
+    python ./scripts/update_jinja_goldens.py
 
   https://github.com/huggingface/transformers/blob/main/src/transformers/utils/chat_template_utils.py
 '''