diff --git a/docs/docs/usage-guide/changing_a_model.md b/docs/docs/usage-guide/changing_a_model.md
index e28bf5e4..b0bde3fa 100644
--- a/docs/docs/usage-guide/changing_a_model.md
+++ b/docs/docs/usage-guide/changing_a_model.md
@@ -234,3 +234,10 @@ To bypass chat templates and temperature controls, set `config.custom_reasoning_
 reasoning_efffort= = "medium" # "low", "medium", "high"
 
 With the OpenAI models that support reasoning effort (eg: o3-mini), you can specify its reasoning effort via `config` section. The default value is `medium`. You can change it to `high` or `low` based on your usage.
+
+### Anthropic models
+
+[config]
+enable_claude_extended_thinking = false # Set to true to enable extended thinking feature
+extended_thinking_budget_tokens = 2048
+extended_thinking_max_output_tokens = 2048
diff --git a/pr_agent/algo/__init__.py b/pr_agent/algo/__init__.py
index 4487b04a..483ea2e0 100644
--- a/pr_agent/algo/__init__.py
+++ b/pr_agent/algo/__init__.py
@@ -60,6 +60,7 @@ MAX_TOKENS = {
     'anthropic/claude-3-5-sonnet-20240620': 100000,
     'anthropic/claude-3-5-sonnet-20241022': 100000,
     'anthropic/claude-3-7-sonnet-20250219': 200000,
+    'claude-3-7-sonnet-20250219': 200000,
     'anthropic/claude-3-5-haiku-20241022': 100000,
     'bedrock/anthropic.claude-instant-v1': 100000,
     'bedrock/anthropic.claude-v2': 100000,
@@ -113,3 +114,8 @@ SUPPORT_REASONING_EFFORT_MODELS = [
     "o3-mini",
     "o3-mini-2025-01-31"
 ]
+
+CLAUDE_EXTENDED_THINKING_MODELS = [
+    "anthropic/claude-3-7-sonnet-20250219",
+    "claude-3-7-sonnet-20250219"
+]
diff --git a/pr_agent/algo/ai_handlers/litellm_ai_handler.py b/pr_agent/algo/ai_handlers/litellm_ai_handler.py
index fe367c9e..af9efc87 100644
--- a/pr_agent/algo/ai_handlers/litellm_ai_handler.py
+++ b/pr_agent/algo/ai_handlers/litellm_ai_handler.py
@@ -6,7 +6,7 @@ import requests
 from litellm import acompletion
 from tenacity import retry, retry_if_exception_type, stop_after_attempt
 
-from pr_agent.algo import NO_SUPPORT_TEMPERATURE_MODELS, SUPPORT_REASONING_EFFORT_MODELS, USER_MESSAGE_ONLY_MODELS
+from pr_agent.algo import CLAUDE_EXTENDED_THINKING_MODELS, NO_SUPPORT_TEMPERATURE_MODELS, SUPPORT_REASONING_EFFORT_MODELS, USER_MESSAGE_ONLY_MODELS
 from pr_agent.algo.ai_handlers.base_ai_handler import BaseAiHandler
 from pr_agent.algo.utils import ReasoningEffort, get_version
 from pr_agent.config_loader import get_settings
@@ -109,6 +109,9 @@ class LiteLLMAIHandler(BaseAiHandler):
         # Models that support reasoning effort
         self.support_reasoning_models = SUPPORT_REASONING_EFFORT_MODELS
 
+        # Models that support extended thinking
+        self.claude_extended_thinking_models = CLAUDE_EXTENDED_THINKING_MODELS
+
     def prepare_logs(self, response, system, user, resp, finish_reason):
         response_log = response.dict().copy()
         response_log['system'] = system
@@ -121,6 +124,43 @@ class LiteLLMAIHandler(BaseAiHandler):
             response_log['main_pr_language'] = 'unknown'
         return response_log
 
+    def _configure_claude_extended_thinking(self, model: str, kwargs: dict) -> dict:
+        """
+        Configure Claude extended thinking parameters if applicable.
+
+        Args:
+            model (str): The AI model being used
+            kwargs (dict): The keyword arguments for the model call
+
+        Returns:
+            dict: Updated kwargs with extended thinking configuration
+        """
+        extended_thinking_budget_tokens = get_settings().config.get("extended_thinking_budget_tokens", 2048)
+        extended_thinking_max_output_tokens = get_settings().config.get("extended_thinking_max_output_tokens", 2048)
+
+        # Validate extended thinking parameters
+        if not isinstance(extended_thinking_budget_tokens, int) or extended_thinking_budget_tokens <= 0:
+            raise ValueError(f"extended_thinking_budget_tokens must be a positive integer, got {extended_thinking_budget_tokens}")
+        if not isinstance(extended_thinking_max_output_tokens, int) or extended_thinking_max_output_tokens <= 0:
+            raise ValueError(f"extended_thinking_max_output_tokens must be a positive integer, got {extended_thinking_max_output_tokens}")
+        if extended_thinking_max_output_tokens < extended_thinking_budget_tokens:
+            raise ValueError(f"extended_thinking_max_output_tokens ({extended_thinking_max_output_tokens}) must be greater than or equal to extended_thinking_budget_tokens ({extended_thinking_budget_tokens})")
+
+        kwargs["thinking"] = {
+            "type": "enabled",
+            "budget_tokens": extended_thinking_budget_tokens
+        }
+        if get_settings().config.verbosity_level >= 2:
+            get_logger().info(f"Adding max output tokens {extended_thinking_max_output_tokens} to model {model}, extended thinking budget tokens: {extended_thinking_budget_tokens}")
+        kwargs["max_tokens"] = extended_thinking_max_output_tokens
+
+        # temperature may only be set to 1 when thinking is enabled
+        if get_settings().config.verbosity_level >= 2:
+            get_logger().info("Temperature may only be set to 1 when thinking is enabled with claude models.")
+        kwargs["temperature"] = 1
+
+        return kwargs
+
     def add_litellm_callbacks(selfs, kwargs) -> dict:
         captured_extra = []
 
@@ -246,6 +286,10 @@ class LiteLLMAIHandler(BaseAiHandler):
                 get_logger().info(f"Adding reasoning_effort with value {reasoning_effort} to model {model}.")
                 kwargs["reasoning_effort"] = reasoning_effort
 
+            # https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking
+            if (model in self.claude_extended_thinking_models) and get_settings().config.get("enable_claude_extended_thinking", False):
+                kwargs = self._configure_claude_extended_thinking(model, kwargs)
+
             if get_settings().litellm.get("enable_callbacks", False):
                 kwargs = self.add_litellm_callbacks(kwargs)
 
@@ -268,13 +312,13 @@ class LiteLLMAIHandler(BaseAiHandler):
                 except json.JSONDecodeError as e:
                     raise ValueError(f"LITELLM.EXTRA_HEADERS contains invalid JSON: {str(e)}")
                 kwargs["extra_headers"] = litellm_extra_headers
-            
+
             get_logger().debug("Prompts", artifact={"system": system, "user": user})
-            
+
             if get_settings().config.verbosity_level >= 2:
                 get_logger().info(f"\nSystem prompt:\n{system}")
                 get_logger().info(f"\nUser prompt:\n{user}")
-                
+
             response = await acompletion(**kwargs)
         except (openai.APIError, openai.APITimeoutError) as e:
             get_logger().warning(f"Error during LLM inference: {e}")
diff --git a/pr_agent/settings/configuration.toml b/pr_agent/settings/configuration.toml
index 5bfb6488..7f8daedf 100644
--- a/pr_agent/settings/configuration.toml
+++ b/pr_agent/settings/configuration.toml
@@ -60,6 +60,10 @@ reasoning_effort = "medium" # "low", "medium", "high"
 enable_auto_approval=false # Set to true to enable auto-approval of PRs under certain conditions
 auto_approve_for_low_review_effort=-1 # -1 to disable, [1-5] to set the threshold for auto-approval
 auto_approve_for_no_suggestions=false # If true, the PR will be auto-approved if there are no suggestions
+# extended thinking for Claude reasoning models
+enable_claude_extended_thinking = false # Set to true to enable extended thinking feature
+extended_thinking_budget_tokens = 2048
+extended_thinking_max_output_tokens = 2048
 
 
 [pr_reviewer] # /review #