diff --git a/tests/integration/llm/providers/test_anthropic.py b/tests/integration/llm/providers/test_anthropic.py
index b5c6af3..7a46517 100644
--- a/tests/integration/llm/providers/test_anthropic.py
+++ b/tests/integration/llm/providers/test_anthropic.py
@@ -37,3 +37,35 @@ def test_anthropic_provider_get_response_live_call():
             wait_time=1,
         )
     )
+
+
+@pytest.mark.integration
+def test_anthropic_provider_web_search():
+    """It invokes web search and returns current information."""
+    from utils.llm.model_registry import (
+        _get_api_key_for_provider,  # type: ignore[import]
+    )
+    from utils.llm.providers.anthropic import AnthropicProvider  # type: ignore[import]
+
+    api_key = _get_api_key_for_provider(AnthropicProvider)
+    assert api_key is not None, "API key should be configured by fixture"
+    provider = anthropic_module.AnthropicProvider(api_key=api_key)
+
+    # Ask a question with a stable, verifiable answer
+    prompt = "What is the official website URL for Python? Just respond with the URL."
+    response = provider.get_response(
+        ANTHROPIC_MODEL,
+        prompt,
+        tools=[
+            {
+                "type": "web_search_20250305",
+                "name": "web_search",
+                "max_uses": 1,
+            }
+        ],
+        max_tokens=256,
+        wait_time=1,
+    )
+
+    assert isinstance(response, str)
+    assert "python.org" in response.lower()
diff --git a/tests/integration/llm/providers/test_openai.py b/tests/integration/llm/providers/test_openai.py
index 7486827..72decba 100644
--- a/tests/integration/llm/providers/test_openai.py
+++ b/tests/integration/llm/providers/test_openai.py
@@ -35,3 +35,29 @@ def test_openai_provider_get_response_live_call():
             wait_time=1,
         )
     )
+
+
+@pytest.mark.integration
+def test_openai_provider_web_search():
+    """It invokes web search and returns current information."""
+    from utils.llm.model_registry import (
+        _get_api_key_for_provider,  # type: ignore[import]
+    )
+    from utils.llm.providers.openai import OpenAIProvider  # type: ignore[import]
+
+    api_key = _get_api_key_for_provider(OpenAIProvider)
+    assert api_key is not None, "API key should be configured by fixture"
+    provider = openai_module.OpenAIProvider(api_key=api_key)
+
+    # Ask a question with a stable, verifiable answer
+    prompt = "What is the official website URL for Python? Just respond with the URL."
+    response = provider.get_response(
+        OPENAI_MODEL,
+        prompt,
+        tools=[{"type": "web_search"}],
+        max_tokens=256,
+        wait_time=1,
+    )
+
+    assert isinstance(response, str)
+    assert "python.org" in response.lower()
diff --git a/utils/llm/providers/anthropic.py b/utils/llm/providers/anthropic.py
index afd9b7f..2bb8892 100644
--- a/utils/llm/providers/anthropic.py
+++ b/utils/llm/providers/anthropic.py
@@ -42,6 +42,7 @@ def _call_model(self, model: "Model", prompt: str, **options: Any) -> str:
         temperature = options.get("temperature")
         max_tokens = options.get("max_tokens")
         assert max_tokens is not None, "max_tokens is required for Anthropic models."
+        tools = options.get("tools")
         model_name = model.full_name
 
         call_args: dict[str, Any] = {
@@ -56,8 +57,15 @@ def _call_model(self, model: "Model", prompt: str, **options: Any) -> str:
         }
         if temperature is not None:
             call_args["temperature"] = temperature
+        if tools is not None:
+            call_args["tools"] = tools
 
         with self._anthropic_console.messages.stream(**call_args) as stream:
             stream.until_done()
 
-        return stream.get_final_message().content[0].text
+        message = stream.get_final_message()
+        # Extract text from the last text block (web search responses have multiple blocks)
+        text_blocks = [block for block in message.content if block.type == "text"]
+        if not text_blocks:
+            return ""
+        return "".join(block.text for block in text_blocks)
diff --git a/utils/llm/providers/openai.py b/utils/llm/providers/openai.py
index 44e1b1a..d2a9a40 100644
--- a/utils/llm/providers/openai.py
+++ b/utils/llm/providers/openai.py
@@ -36,25 +36,25 @@ def __init__(self, *, api_key: str | None = None, default_wait_time: int | None
         self._openai_client = OpenAI(api_key=api_key)
 
     def _call_model(self, model: "Model", prompt: str, **options: Any) -> str:
-        temperature = options.get("temperature", 0.8)
-        max_tokens = options.get("max_tokens")
         model_name = model.full_name
 
-        # OpenAI doesn't support temperature for reasoning models
-        if model.reasoning_model:
-            request_payload: Dict[str, Any] = {
-                "model": model_name,
-                "input": prompt,
-            }
-        else:
-            request_payload = {
-                "model": model_name,
-                "input": prompt,
-                "temperature": temperature,
-            }
-
-        if max_tokens is not None:
-            request_payload["max_output_tokens"] = max_tokens
+        request_payload: Dict[str, Any] = {
+            "model": model_name,
+            "input": prompt,
+        }
+
+        # Add default temperature for non-reasoning models (unless explicitly provided)
+        if not model.reasoning_model and "temperature" not in options:
+            request_payload["temperature"] = 0.8
+
+        # Pass through all options, with special handling for some keys
+        for key, value in options.items():
+            if key == "max_tokens":
+                request_payload["max_output_tokens"] = value
+            elif key == "temperature" and model.reasoning_model:
+                continue  # OpenAI doesn't support temperature for reasoning models
+            else:
+                request_payload[key] = value
 
         response = self._openai_client.responses.create(**request_payload)