diff --git a/src/mcp_agent/llm/providers/augmented_llm_deepseek.py b/src/mcp_agent/llm/providers/augmented_llm_deepseek.py
index 4e6b5392b..26c3ed664 100644
--- a/src/mcp_agent/llm/providers/augmented_llm_deepseek.py
+++ b/src/mcp_agent/llm/providers/augmented_llm_deepseek.py
@@ -1,6 +1,10 @@
+from typing import List, Tuple, Type
+
 from mcp_agent.core.request_params import RequestParams
 from mcp_agent.llm.provider_types import Provider
 from mcp_agent.llm.providers.augmented_llm_openai import OpenAIAugmentedLLM
+from mcp_agent.mcp.interfaces import ModelT
+from mcp_agent.mcp.prompt_message_multipart import PromptMessageMultipart
 
 DEEPSEEK_BASE_URL = "https://api.deepseek.com"
 DEFAULT_DEEPSEEK_MODEL = "deepseekchat"  # current Deepseek only has two type models
@@ -20,6 +24,9 @@ def _initialize_default_params(self, kwargs: dict) -> RequestParams:
             parallel_tool_calls=True,
             max_iterations=10,
             use_history=True,
+            response_format={
+                                "type": "json_object"
+                            }
         )
 
     def _base_url(self) -> str:
@@ -28,3 +35,46 @@ def _base_url(self) -> str:
             base_url = self.context.config.deepseek.base_url
 
         return base_url if base_url else DEEPSEEK_BASE_URL
+    
+    async def _apply_prompt_provider_specific_structured(
+        self,
+        multipart_messages: List[PromptMessageMultipart],
+        model: Type[ModelT],
+        request_params: RequestParams | None = None,
+    ) -> Tuple[ModelT | None, PromptMessageMultipart]:  # noqa: F821
+        request_params = self.get_request_params(request_params)
+
+        # Get the full schema and extract just the properties
+        full_schema = model.model_json_schema()
+        properties = full_schema.get("properties", {})
+        required_fields = full_schema.get("required", [])
+        
+        # Create a cleaner format description
+        format_description = "{\n"
+        for field_name, field_info in properties.items():
+            field_type = field_info.get("type", "string")
+            description = field_info.get("description", "")
+            format_description += f'  "{field_name}": "{field_type}"'
+            if description:
+                format_description += f'  // {description}'
+            if field_name in required_fields:
+                format_description += '  // REQUIRED'
+            format_description += '\n'
+        format_description += "}"
+
+        multipart_messages[-1].add_text(
+            f"""YOU MUST RESPOND WITH A JSON OBJECT IN EXACTLY THIS FORMAT:
+            {format_description}
+
+            IMPORTANT RULES:
+            - Respond ONLY with the JSON object, no other text
+            - Do NOT include "properties" or "schema" wrappers
+            - Do NOT use code fences or markdown
+            - The response must be valid JSON that matches the format above
+            - All required fields must be included"""
+        )
+
+        result: PromptMessageMultipart = await self._apply_prompt_provider_specific(
+            multipart_messages, request_params
+        )
+        return self._structured_from_multipart(result, model)
diff --git a/tests/e2e/structured/test_structured_outputs.py b/tests/e2e/structured/test_structured_outputs.py
index d768f1eec..30fd05f72 100644
--- a/tests/e2e/structured/test_structured_outputs.py
+++ b/tests/e2e/structured/test_structured_outputs.py
@@ -22,8 +22,9 @@ class FormattedResponse(BaseModel):
 @pytest.mark.parametrize(
     "model_name",
     [
-        "generic.qwen2.5:latest",
-        "generic.llama3.2:latest",
+        #        "generic.qwen2.5:latest",
+        #       "generic.llama3.2:latest",
+        "deepseek-chat",
         "haiku",
         "sonnet",
         "gpt-4.1",