diff --git a/mlx_lm/server.py b/mlx_lm/server.py
index ecfaa65..37b4165 100644
--- a/mlx_lm/server.py
+++ b/mlx_lm/server.py
@@ -524,6 +524,13 @@ class ResponseGenerator:
 
             if tokenizer.has_chat_template:
                 process_message_content(messages)
+                if tools and not tokenizer.has_tool_calling:
+                    logging.warning(
+                        "Received tools but model does not support tool calling. "
+                        "If you think this is an error, file an issue here: "
+                        "https://github.com/ml-explore/mlx-lm/issues"
+                    )
+
                 return tokenizer.apply_chat_template(
                     messages,
                     tools=tools,
diff --git a/mlx_lm/tokenizer_utils.py b/mlx_lm/tokenizer_utils.py
index 2aa2e21..19fdae2 100644
--- a/mlx_lm/tokenizer_utils.py
+++ b/mlx_lm/tokenizer_utils.py
@@ -1,13 +1,12 @@
 import importlib
 import json
+import warnings
 from functools import partial
 from json import JSONDecodeError
 from typing import Any, Dict, List, Optional
 
 from transformers import AutoTokenizer, PreTrainedTokenizerFast
 
-from .tool_parsers.json_tools import parse_tool_call as default_tool_parser
-
 
 class StreamingDetokenizer:
     """The streaming detokenizer interface so that we can detokenize one token at a time.
@@ -276,7 +275,7 @@ class TokenizerWrapper:
         self.has_chat_template = (
             tokenizer.chat_template is not None or chat_template is not None
         )
-        self._tool_parser = tool_parser or default_tool_parser
+        self._tool_parser = tool_parser
         self._tool_call_start = tool_call_start
         self._tool_call_end = tool_call_end
 
@@ -290,11 +289,13 @@ class TokenizerWrapper:
                 self._think_end_id = vocab[think_end]
                 break
 
-        # Fallback to defaults if no tool call tokens are provided
-        if tool_call_start and tool_call_start not in vocab:
-            raise ValueError("Tool call start token not in vocab")
-        if tool_call_end and tool_call_end not in vocab:
-            raise ValueError("Tool call end token not in vocab")
+        # Disable tool calling if tool call tokens aren't in vocab
+        if (tool_call_start and tool_call_start not in vocab) or (
+            tool_call_end and tool_call_end not in vocab
+        ):
+            self._tool_call_start = None
+            self._tool_call_end = None
+            self._tool_parser = None
 
     def apply_chat_template(self, *args, tokenize=True, **kwargs):
         if self._chat_template is not None: