Lorenze/ensure hooks work with lite agents flows (#3981)

* liteagent support hooks

* wip llm.call hooks work - needs tests for this

* fix tests

* fixed more

* more tool hooks test cassettes

lib/crewai/src/crewai/hooks/llm_hooks.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-from typing import TYPE_CHECKING
+from typing import TYPE_CHECKING, Any, cast
 
 from crewai.events.event_listener import event_listener
 from crewai.hooks.types import AfterLLMCallHookType, BeforeLLMCallHookType
@@ -9,17 +9,22 @@ from crewai.utilities.printer import Printer
 
 if TYPE_CHECKING:
     from crewai.agents.crew_agent_executor import CrewAgentExecutor
+    from crewai.lite_agent import LiteAgent
+    from crewai.llms.base_llm import BaseLLM
+    from crewai.utilities.types import LLMMessage
 
 
 class LLMCallHookContext:
-    """Context object passed to LLM call hooks with full executor access.
+    """Context object passed to LLM call hooks.
 
-    Provides hooks with complete access to the executor state, allowing
+    Provides hooks with complete access to the execution state, allowing
     modification of messages, responses, and executor attributes.
 
+    Supports both executor-based calls (agents in crews/flows) and direct LLM calls.
+
     Attributes:
-        executor: Full reference to the CrewAgentExecutor instance
-        messages: Direct reference to executor.messages (mutable list).
+        executor: Reference to the executor (CrewAgentExecutor/LiteAgent) or None for direct calls
+        messages: Direct reference to messages (mutable list).
             Can be modified in both before_llm_call and after_llm_call hooks.
             Modifications in after_llm_call hooks persist to the next iteration,
             allowing hooks to modify conversation history for subsequent LLM calls.
@@ -27,33 +32,75 @@ class LLMCallHookContext:
             Do NOT replace the list (e.g., context.messages = []), as this will break
             the executor. Use context.messages.append() or context.messages.extend()
             instead of assignment.
-        agent: Reference to the agent executing the task
-        task: Reference to the task being executed
-        crew: Reference to the crew instance
+        agent: Reference to the agent executing the task (None for direct LLM calls)
+        task: Reference to the task being executed (None for direct LLM calls or LiteAgent)
+        crew: Reference to the crew instance (None for direct LLM calls or LiteAgent)
         llm: Reference to the LLM instance
-        iterations: Current iteration count
+        iterations: Current iteration count (0 for direct LLM calls)
         response: LLM response string (only set for after_llm_call hooks).
             Can be modified by returning a new string from after_llm_call hook.
     """
 
+    executor: CrewAgentExecutor | LiteAgent | None
+    messages: list[LLMMessage]
+    agent: Any
+    task: Any
+    crew: Any
+    llm: BaseLLM | None | str | Any
+    iterations: int
+    response: str | None
+
     def __init__(
         self,
-        executor: CrewAgentExecutor,
+        executor: CrewAgentExecutor | LiteAgent | None = None,
         response: str | None = None,
+        messages: list[LLMMessage] | None = None,
+        llm: BaseLLM | str | Any | None = None,  # TODO: look into
+        agent: Any | None = None,
+        task: Any | None = None,
+        crew: Any | None = None,
     ) -> None:
-        """Initialize hook context with executor reference.
+        """Initialize hook context with executor reference or direct parameters.
 
         Args:
-            executor: The CrewAgentExecutor instance
+            executor: The CrewAgentExecutor or LiteAgent instance (None for direct LLM calls)
             response: Optional response string (for after_llm_call hooks)
+            messages: Optional messages list (for direct LLM calls when executor is None)
+            llm: Optional LLM instance (for direct LLM calls when executor is None)
+            agent: Optional agent reference (for direct LLM calls when executor is None)
+            task: Optional task reference (for direct LLM calls when executor is None)
+            crew: Optional crew reference (for direct LLM calls when executor is None)
         """
-        self.executor = executor
-        self.messages = executor.messages
-        self.agent = executor.agent
-        self.task = executor.task
-        self.crew = executor.crew
-        self.llm = executor.llm
-        self.iterations = executor.iterations
+        if executor is not None:
+            # Existing path: extract from executor
+            self.executor = executor
+            self.messages = executor.messages
+            self.llm = executor.llm
+            self.iterations = executor.iterations
+            # Handle CrewAgentExecutor vs LiteAgent differences
+            if hasattr(executor, "agent"):
+                self.agent = executor.agent
+                self.task = cast("CrewAgentExecutor", executor).task
+                self.crew = cast("CrewAgentExecutor", executor).crew
+            else:
+                # LiteAgent case - is the agent itself, doesn't have task/crew
+                self.agent = (
+                    executor.original_agent
+                    if hasattr(executor, "original_agent")
+                    else executor
+                )
+                self.task = None
+                self.crew = None
+        else:
+            # New path: direct LLM call with explicit parameters
+            self.executor = None
+            self.messages = messages or []
+            self.llm = llm
+            self.agent = agent
+            self.task = task
+            self.crew = crew
+            self.iterations = 0
+
         self.response = response
 
     def request_human_input(

lib/crewai/src/crewai/lite_agent.py

@@ -38,6 +38,8 @@ from crewai.events.types.agent_events import (
 )
 from crewai.events.types.logging_events import AgentLogsExecutionEvent
 from crewai.flow.flow_trackable import FlowTrackable
+from crewai.hooks.llm_hooks import get_after_llm_call_hooks, get_before_llm_call_hooks
+from crewai.hooks.types import AfterLLMCallHookType, BeforeLLMCallHookType
 from crewai.lite_agent_output import LiteAgentOutput
 from crewai.llm import LLM
 from crewai.llms.base_llm import BaseLLM
@@ -155,6 +157,12 @@ class LiteAgent(FlowTrackable, BaseModel):
     _guardrail: GuardrailCallable | None = PrivateAttr(default=None)
     _guardrail_retry_count: int = PrivateAttr(default=0)
     _callbacks: list[TokenCalcHandler] = PrivateAttr(default_factory=list)
+    _before_llm_call_hooks: list[BeforeLLMCallHookType] = PrivateAttr(
+        default_factory=get_before_llm_call_hooks
+    )
+    _after_llm_call_hooks: list[AfterLLMCallHookType] = PrivateAttr(
+        default_factory=get_after_llm_call_hooks
+    )
 
     @model_validator(mode="after")
     def setup_llm(self) -> Self:
@@ -246,6 +254,26 @@ class LiteAgent(FlowTrackable, BaseModel):
         """Return the original role for compatibility with tool interfaces."""
         return self.role
 
+    @property
+    def before_llm_call_hooks(self) -> list[BeforeLLMCallHookType]:
+        """Get the before_llm_call hooks for this agent."""
+        return self._before_llm_call_hooks
+
+    @property
+    def after_llm_call_hooks(self) -> list[AfterLLMCallHookType]:
+        """Get the after_llm_call hooks for this agent."""
+        return self._after_llm_call_hooks
+
+    @property
+    def messages(self) -> list[LLMMessage]:
+        """Get the messages list for hook context compatibility."""
+        return self._messages
+
+    @property
+    def iterations(self) -> int:
+        """Get the current iteration count for hook context compatibility."""
+        return self._iterations
+
     def kickoff(
         self,
         messages: str | list[LLMMessage],
@@ -504,7 +532,7 @@ class LiteAgent(FlowTrackable, BaseModel):
             AgentFinish: The final result of the agent execution.
         """
         # Execute the agent loop
-        formatted_answer = None
+        formatted_answer: AgentAction | AgentFinish | None = None
         while not isinstance(formatted_answer, AgentFinish):
             try:
                 if has_reached_max_iterations(self._iterations, self.max_iterations):
@@ -526,6 +554,7 @@ class LiteAgent(FlowTrackable, BaseModel):
                         callbacks=self._callbacks,
                         printer=self._printer,
                         from_agent=self,
+                        executor_context=self,
                     )
 
                 except Exception as e:

lib/crewai/src/crewai/llm.py

@@ -1642,6 +1642,10 @@ class LLM(BaseLLM):
                 if message.get("role") == "system":
                     msg_role: Literal["assistant"] = "assistant"
                     message["role"] = msg_role
+
+        if not self._invoke_before_llm_call_hooks(messages, from_agent):
+            raise ValueError("LLM call blocked by before_llm_call hook")
+
         # --- 5) Set up callbacks if provided
         with suppress_warnings():
             if callbacks and len(callbacks) > 0:
@@ -1651,7 +1655,16 @@ class LLM(BaseLLM):
                 params = self._prepare_completion_params(messages, tools)
                 # --- 7) Make the completion call and handle response
                 if self.stream:
-                    return self._handle_streaming_response(
+                    result = self._handle_streaming_response(
+                        params=params,
+                        callbacks=callbacks,
+                        available_functions=available_functions,
+                        from_task=from_task,
+                        from_agent=from_agent,
+                        response_model=response_model,
+                    )
+                else:
+                    result = self._handle_non_streaming_response(
                         params=params,
                         callbacks=callbacks,
                         available_functions=available_functions,
@@ -1660,14 +1673,12 @@ class LLM(BaseLLM):
                         response_model=response_model,
                     )
 
-                return self._handle_non_streaming_response(
-                    params=params,
-                    callbacks=callbacks,
-                    available_functions=available_functions,
-                    from_task=from_task,
-                    from_agent=from_agent,
-                    response_model=response_model,
-                )
+                if isinstance(result, str):
+                    result = self._invoke_after_llm_call_hooks(
+                        messages, result, from_agent
+                    )
+
+                return result
             except LLMContextLengthExceededError:
                 # Re-raise LLMContextLengthExceededError as it should be handled
                 # by the CrewAgentExecutor._invoke_loop method, which can then decide

lib/crewai/src/crewai/llms/base_llm.py

@@ -314,7 +314,7 @@ class BaseLLM(ABC):
         call_type: LLMCallType,
         from_task: Task | None = None,
         from_agent: Agent | None = None,
-        messages: str | list[dict[str, Any]] | None = None,
+        messages: str | list[LLMMessage] | None = None,
     ) -> None:
         """Emit LLM call completed event."""
         crewai_event_bus.emit(
@@ -586,3 +586,134 @@ class BaseLLM(ABC):
             Dictionary with token usage totals
         """
         return UsageMetrics(**self._token_usage)
+
+    def _invoke_before_llm_call_hooks(
+        self,
+        messages: list[LLMMessage],
+        from_agent: Agent | None = None,
+    ) -> bool:
+        """Invoke before_llm_call hooks for direct LLM calls (no agent context).
+
+        This method should be called by native provider implementations before
+        making the actual LLM call when from_agent is None (direct calls).
+
+        Args:
+            messages: The messages being sent to the LLM
+            from_agent: The agent making the call (None for direct calls)
+
+        Returns:
+            True if LLM call should proceed, False if blocked by hook
+
+        Example:
+            >>> # In a native provider's call() method:
+            >>> if from_agent is None and not self._invoke_before_llm_call_hooks(
+            ...     messages, from_agent
+            ... ):
+            ...     raise ValueError("LLM call blocked by hook")
+        """
+        # Only invoke hooks for direct calls (no agent context)
+        if from_agent is not None:
+            return True
+
+        from crewai.hooks.llm_hooks import (
+            LLMCallHookContext,
+            get_before_llm_call_hooks,
+        )
+        from crewai.utilities.printer import Printer
+
+        before_hooks = get_before_llm_call_hooks()
+        if not before_hooks:
+            return True
+
+        hook_context = LLMCallHookContext(
+            executor=None,
+            messages=messages,
+            llm=self,
+            agent=None,
+            task=None,
+            crew=None,
+        )
+        printer = Printer()
+
+        try:
+            for hook in before_hooks:
+                result = hook(hook_context)
+                if result is False:
+                    printer.print(
+                        content="LLM call blocked by before_llm_call hook",
+                        color="yellow",
+                    )
+                    return False
+        except Exception as e:
+            printer.print(
+                content=f"Error in before_llm_call hook: {e}",
+                color="yellow",
+            )
+
+        return True
+
+    def _invoke_after_llm_call_hooks(
+        self,
+        messages: list[LLMMessage],
+        response: str,
+        from_agent: Agent | None = None,
+    ) -> str:
+        """Invoke after_llm_call hooks for direct LLM calls (no agent context).
+
+        This method should be called by native provider implementations after
+        receiving the LLM response when from_agent is None (direct calls).
+
+        Args:
+            messages: The messages that were sent to the LLM
+            response: The response from the LLM
+            from_agent: The agent that made the call (None for direct calls)
+
+        Returns:
+            The potentially modified response string
+
+        Example:
+            >>> # In a native provider's call() method:
+            >>> if from_agent is None and isinstance(result, str):
+            ...     result = self._invoke_after_llm_call_hooks(
+            ...         messages, result, from_agent
+            ...     )
+        """
+        # Only invoke hooks for direct calls (no agent context)
+        if from_agent is not None or not isinstance(response, str):
+            return response
+
+        from crewai.hooks.llm_hooks import (
+            LLMCallHookContext,
+            get_after_llm_call_hooks,
+        )
+        from crewai.utilities.printer import Printer
+
+        after_hooks = get_after_llm_call_hooks()
+        if not after_hooks:
+            return response
+
+        hook_context = LLMCallHookContext(
+            executor=None,
+            messages=messages,
+            llm=self,
+            agent=None,
+            task=None,
+            crew=None,
+            response=response,
+        )
+        printer = Printer()
+        modified_response = response
+
+        try:
+            for hook in after_hooks:
+                result = hook(hook_context)
+                if result is not None and isinstance(result, str):
+                    modified_response = result
+                    hook_context.response = modified_response
+        except Exception as e:
+            printer.print(
+                content=f"Error in after_llm_call hook: {e}",
+                color="yellow",
+            )
+
+        return modified_response

lib/crewai/src/crewai/llms/providers/anthropic/completion.py

@@ -187,6 +187,9 @@ class AnthropicCompletion(BaseLLM):
                 messages
             )
 
+            if not self._invoke_before_llm_call_hooks(formatted_messages, from_agent):
+                raise ValueError("LLM call blocked by before_llm_call hook")
+
             # Prepare completion parameters
             completion_params = self._prepare_completion_params(
                 formatted_messages, system_message, tools
@@ -494,7 +497,9 @@ class AnthropicCompletion(BaseLLM):
         if usage.get("total_tokens", 0) > 0:
             logging.info(f"Anthropic API usage: {usage}")
 
-        return content
+        return self._invoke_after_llm_call_hooks(
+            params["messages"], content, from_agent
+        )
 
     def _handle_streaming_completion(
         self,
@@ -588,7 +593,9 @@ class AnthropicCompletion(BaseLLM):
             messages=params["messages"],
         )
 
-        return full_response
+        return self._invoke_after_llm_call_hooks(
+            params["messages"], full_response, from_agent
+        )
 
     def _handle_tool_use_conversation(
         self,

lib/crewai/src/crewai/llms/providers/azure/completion.py

@@ -216,6 +216,9 @@ class AzureCompletion(BaseLLM):
             # Format messages for Azure
             formatted_messages = self._format_messages_for_azure(messages)
 
+            if not self._invoke_before_llm_call_hooks(formatted_messages, from_agent):
+                raise ValueError("LLM call blocked by before_llm_call hook")
+
             # Prepare completion parameters
             completion_params = self._prepare_completion_params(
                 formatted_messages, tools, response_model
@@ -550,6 +553,10 @@ class AzureCompletion(BaseLLM):
                 messages=params["messages"],
             )
 
+            content = self._invoke_after_llm_call_hooks(
+                params["messages"], content, from_agent
+            )
+
         except Exception as e:
             if is_context_length_exceeded(e):
                 logging.error(f"Context window exceeded: {e}")
@@ -642,7 +649,9 @@ class AzureCompletion(BaseLLM):
             messages=params["messages"],
         )
 
-        return full_response
+        return self._invoke_after_llm_call_hooks(
+            params["messages"], full_response, from_agent
+        )
 
     async def _ahandle_completion(
         self,

lib/crewai/src/crewai/llms/providers/bedrock/completion.py

@@ -312,9 +312,14 @@ class BedrockCompletion(BaseLLM):
 
             # Format messages for Converse API
             formatted_messages, system_message = self._format_messages_for_converse(
-                messages  # type: ignore[arg-type]
+                messages
             )
 
+            if not self._invoke_before_llm_call_hooks(
+                cast(list[LLMMessage], formatted_messages), from_agent
+            ):
+                raise ValueError("LLM call blocked by before_llm_call hook")
+
             # Prepare request body
             body: BedrockConverseRequestBody = {
                 "inferenceConfig": self._get_inference_config(),
@@ -356,11 +361,19 @@ class BedrockCompletion(BaseLLM):
 
             if self.stream:
                 return self._handle_streaming_converse(
-                    formatted_messages, body, available_functions, from_task, from_agent
+                    cast(list[LLMMessage], formatted_messages),
+                    body,
+                    available_functions,
+                    from_task,
+                    from_agent,
                 )
 
             return self._handle_converse(
-                formatted_messages, body, available_functions, from_task, from_agent
+                cast(list[LLMMessage], formatted_messages),
+                body,
+                available_functions,
+                from_task,
+                from_agent,
             )
 
         except Exception as e:
@@ -481,7 +494,7 @@ class BedrockCompletion(BaseLLM):
 
     def _handle_converse(
         self,
-        messages: list[dict[str, Any]],
+        messages: list[LLMMessage],
         body: BedrockConverseRequestBody,
         available_functions: Mapping[str, Any] | None = None,
         from_task: Any | None = None,
@@ -605,7 +618,11 @@ class BedrockCompletion(BaseLLM):
                 messages=messages,
             )
 
-            return text_content
+            return self._invoke_after_llm_call_hooks(
+                messages,
+                text_content,
+                from_agent,
+            )
 
         except ClientError as e:
             # Handle all AWS ClientError exceptions as per documentation
@@ -662,7 +679,7 @@ class BedrockCompletion(BaseLLM):
 
     def _handle_streaming_converse(
         self,
-        messages: list[dict[str, Any]],
+        messages: list[LLMMessage],
         body: BedrockConverseRequestBody,
         available_functions: dict[str, Any] | None = None,
         from_task: Any | None = None,
@@ -1149,16 +1166,25 @@ class BedrockCompletion(BaseLLM):
             messages=messages,
         )
 
-        return full_response
+        return self._invoke_after_llm_call_hooks(
+            messages,
+            full_response,
+            from_agent,
+        )
 
     def _format_messages_for_converse(
-        self, messages: str | list[dict[str, str]]
+        self, messages: str | list[LLMMessage]
     ) -> tuple[list[dict[str, Any]], str | None]:
-        """Format messages for Converse API following AWS documentation."""
-        # Use base class formatting first
-        formatted_messages = self._format_messages(messages)  # type: ignore[arg-type]
+        """Format messages for Converse API following AWS documentation.
 
-        converse_messages = []
+        Note: Returns dict[str, Any] instead of LLMMessage because Bedrock uses
+        a different content structure: {"role": str, "content": [{"text": str}]}
+        rather than the standard {"role": str, "content": str}.
+        """
+        # Use base class formatting first
+        formatted_messages = self._format_messages(messages)
+
+        converse_messages: list[dict[str, Any]] = []
         system_message: str | None = None
 
         for message in formatted_messages:

lib/crewai/src/crewai/llms/providers/gemini/completion.py

@@ -246,6 +246,11 @@ class GeminiCompletion(BaseLLM):
                 messages
             )
 
+            messages_for_hooks = self._convert_contents_to_dict(formatted_content)
+
+            if not self._invoke_before_llm_call_hooks(messages_for_hooks, from_agent):
+                raise ValueError("LLM call blocked by before_llm_call hook")
+
             config = self._prepare_generation_config(
                 system_instruction, tools, response_model
             )
@@ -559,7 +564,9 @@ class GeminiCompletion(BaseLLM):
             messages=messages_for_event,
         )
 
-        return content
+        return self._invoke_after_llm_call_hooks(
+            messages_for_event, content, from_agent
+        )
 
     def _handle_streaming_completion(
         self,
@@ -639,7 +646,9 @@ class GeminiCompletion(BaseLLM):
             messages=messages_for_event,
         )
 
-        return full_response
+        return self._invoke_after_llm_call_hooks(
+            messages_for_event, full_response, from_agent
+        )
 
     async def _ahandle_completion(
         self,
@@ -787,7 +796,159 @@ class GeminiCompletion(BaseLLM):
             messages=messages_for_event,
         )
 
-        return full_response
+        return self._invoke_after_llm_call_hooks(
+            messages_for_event, full_response, from_agent
+        )
+
+    async def _ahandle_completion(
+        self,
+        contents: list[types.Content],
+        system_instruction: str | None,
+        config: types.GenerateContentConfig,
+        available_functions: dict[str, Any] | None = None,
+        from_task: Any | None = None,
+        from_agent: Any | None = None,
+        response_model: type[BaseModel] | None = None,
+    ) -> str | Any:
+        """Handle async non-streaming content generation."""
+        try:
+            # The API accepts list[Content] but mypy is overly strict about variance
+            contents_for_api: Any = contents
+            response = await self.client.aio.models.generate_content(
+                model=self.model,
+                contents=contents_for_api,
+                config=config,
+            )
+
+            usage = self._extract_token_usage(response)
+        except Exception as e:
+            if is_context_length_exceeded(e):
+                logging.error(f"Context window exceeded: {e}")
+                raise LLMContextLengthExceededError(str(e)) from e
+            raise e from e
+
+        self._track_token_usage_internal(usage)
+
+        if response.candidates and (self.tools or available_functions):
+            candidate = response.candidates[0]
+            if candidate.content and candidate.content.parts:
+                for part in candidate.content.parts:
+                    if hasattr(part, "function_call") and part.function_call:
+                        function_name = part.function_call.name
+                        if function_name is None:
+                            continue
+                        function_args = (
+                            dict(part.function_call.args)
+                            if part.function_call.args
+                            else {}
+                        )
+
+                        result = self._handle_tool_execution(
+                            function_name=function_name,
+                            function_args=function_args,
+                            available_functions=available_functions or {},
+                            from_task=from_task,
+                            from_agent=from_agent,
+                        )
+
+                        if result is not None:
+                            return result
+
+        content = response.text or ""
+        content = self._apply_stop_words(content)
+
+        messages_for_event = self._convert_contents_to_dict(contents)
+
+        self._emit_call_completed_event(
+            response=content,
+            call_type=LLMCallType.LLM_CALL,
+            from_task=from_task,
+            from_agent=from_agent,
+            messages=messages_for_event,
+        )
+
+        return content
+
+    async def _ahandle_streaming_completion(
+        self,
+        contents: list[types.Content],
+        config: types.GenerateContentConfig,
+        available_functions: dict[str, Any] | None = None,
+        from_task: Any | None = None,
+        from_agent: Any | None = None,
+        response_model: type[BaseModel] | None = None,
+    ) -> str:
+        """Handle async streaming content generation."""
+        full_response = ""
+        function_calls: dict[str, dict[str, Any]] = {}
+
+        # The API accepts list[Content] but mypy is overly strict about variance
+        contents_for_api: Any = contents
+        stream = await self.client.aio.models.generate_content_stream(
+            model=self.model,
+            contents=contents_for_api,
+            config=config,
+        )
+        async for chunk in stream:
+            if chunk.text:
+                full_response += chunk.text
+                self._emit_stream_chunk_event(
+                    chunk=chunk.text,
+                    from_task=from_task,
+                    from_agent=from_agent,
+                )
+
+            if chunk.candidates:
+                candidate = chunk.candidates[0]
+                if candidate.content and candidate.content.parts:
+                    for part in candidate.content.parts:
+                        if hasattr(part, "function_call") and part.function_call:
+                            call_id = part.function_call.name or "default"
+                            if call_id not in function_calls:
+                                function_calls[call_id] = {
+                                    "name": part.function_call.name,
+                                    "args": dict(part.function_call.args)
+                                    if part.function_call.args
+                                    else {},
+                                }
+
+        if function_calls and available_functions:
+            for call_data in function_calls.values():
+                function_name = call_data["name"]
+                function_args = call_data["args"]
+
+                # Skip if function_name is None
+                if not isinstance(function_name, str):
+                    continue
+
+                # Ensure function_args is a dict
+                if not isinstance(function_args, dict):
+                    function_args = {}
+
+                result = self._handle_tool_execution(
+                    function_name=function_name,
+                    function_args=function_args,
+                    available_functions=available_functions,
+                    from_task=from_task,
+                    from_agent=from_agent,
+                )
+
+                if result is not None:
+                    return result
+
+        messages_for_event = self._convert_contents_to_dict(contents)
+
+        self._emit_call_completed_event(
+            response=full_response,
+            call_type=LLMCallType.LLM_CALL,
+            from_task=from_task,
+            from_agent=from_agent,
+            messages=messages_for_event,
+        )
+
+        return self._invoke_after_llm_call_hooks(
+            messages_for_event, full_response, from_agent
+        )
 
     def supports_function_calling(self) -> bool:
         """Check if the model supports function calling."""
@@ -851,7 +1012,7 @@ class GeminiCompletion(BaseLLM):
     def _convert_contents_to_dict(
         self,
         contents: list[types.Content],
-    ) -> list[dict[str, str]]:
+    ) -> list[LLMMessage]:
         """Convert contents to dict format."""
         result: list[dict[str, str]] = []
         for content_obj in contents:

lib/crewai/src/crewai/llms/providers/openai/completion.py

@@ -190,6 +190,9 @@ class OpenAICompletion(BaseLLM):
 
             formatted_messages = self._format_messages(messages)
 
+            if not self._invoke_before_llm_call_hooks(formatted_messages, from_agent):
+                raise ValueError("LLM call blocked by before_llm_call hook")
+
             completion_params = self._prepare_completion_params(
                 messages=formatted_messages, tools=tools
             )
@@ -474,6 +477,10 @@ class OpenAICompletion(BaseLLM):
 
             if usage.get("total_tokens", 0) > 0:
                 logging.info(f"OpenAI API usage: {usage}")
+
+            content = self._invoke_after_llm_call_hooks(
+                params["messages"], content, from_agent
+            )
         except NotFoundError as e:
             error_msg = f"Model {self.model} not found: {e}"
             logging.error(error_msg)
@@ -629,7 +636,9 @@ class OpenAICompletion(BaseLLM):
             messages=params["messages"],
         )
 
-        return full_response
+        return self._invoke_after_llm_call_hooks(
+            params["messages"], full_response, from_agent
+        )
 
     async def _ahandle_completion(
         self,

lib/crewai/src/crewai/utilities/agent_utils.py

@@ -237,7 +237,7 @@ def get_llm_response(
     from_task: Task | None = None,
     from_agent: Agent | LiteAgent | None = None,
     response_model: type[BaseModel] | None = None,
-    executor_context: CrewAgentExecutor | None = None,
+    executor_context: CrewAgentExecutor | LiteAgent | None = None,
 ) -> str:
     """Call the LLM and return the response, handling any invalid responses.
 
@@ -727,7 +727,7 @@ def load_agent_from_repository(from_repository: str) -> dict[str, Any]:
 
 
 def _setup_before_llm_call_hooks(
-    executor_context: CrewAgentExecutor | None, printer: Printer
+    executor_context: CrewAgentExecutor | LiteAgent | None, printer: Printer
 ) -> bool:
     """Setup and invoke before_llm_call hooks for the executor context.
 
@@ -777,7 +777,7 @@ def _setup_before_llm_call_hooks(
 
 
 def _setup_after_llm_call_hooks(
-    executor_context: CrewAgentExecutor | None,
+    executor_context: CrewAgentExecutor | LiteAgent | None,
     answer: str,
     printer: Printer,
 ) -> str:

lib/crewai/tests/cassettes/TestLLMHooksIntegration.test_direct_llm_call_hooks_integration.yaml

@@ -0,0 +1,82 @@
+interactions:
+- request:
+    body: '{"messages":[{"role":"user","content":"Say hello"}],"model":"gpt-4o-mini"}'
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate, zstd
+      authorization:
+      - AUTHORIZATION-XXX
+      connection:
+      - keep-alive
+      content-length:
+      - '74'
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.109.1
+      x-stainless-arch:
+      - arm64
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - MacOS
+      x-stainless-package-version:
+      - 1.109.1
+      x-stainless-read-timeout:
+      - '600'
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.13.3
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAAwAAAP//jFJNb9QwEL3nVww+b9Am7Ee7FyT2wCIQ0CJxqarItSdZg+Ox7AmwVPvf
+        KyftJv1A4uLDvHnP783MbQYgjBYbEGovWbXe5tvm6/bv5ZeDu5AmlubTzr///G778fKi+O6/iVli
+        0M0PVPzAeq2o9RbZkBtgFVAyJtVivVqUZbku3vRASxptojWe8wXlrXEmL+flIp+v8+Lsnr0nozCK
+        DVxlAAC3/Zt8Oo1/xAbms4dKizHKBsXm1AQgAtlUETJGE1k6FrMRVOQYXW99h9bSK9jRb1DSwQcY
+        CHCgDpi0PLydEgPWXZTJvOusnQDSOWKZwveWr++R48mkpcYHuolPqKI2zsR9FVBGcslQZPKiR48Z
+        wHU/jO5RPuEDtZ4rpp/Yf3c+qIlxA88xJpZ2LBdnsxe0Ko0sjY2TUQol1R71yBznLjttaAJkk8TP
+        vbykPaQ2rvkf+RFQCj2jrnxAbdTjvGNbwHSe/2o7Tbg3LCKGX0ZhxQZD2oLGWnZ2OBoRD5GxrWrj
+        Ggw+mOFyal8tV3NZr3C5PBfZMbsDAAD//wMARXm1qUcDAAA=
+    headers:
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Thu, 27 Nov 2025 05:51:54 GMT
+      Server:
+      - cloudflare
+      Strict-Transport-Security:
+      - max-age=31536000; includeSubDomains; preload
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - nosniff
+      access-control-expose-headers:
+      - X-Request-ID
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      openai-version:
+      - '2020-10-01'
+      x-openai-proxy-wasm:
+      - v0.1
+    status:
+      code: 200
+      message: OK
+version: 1

lib/crewai/tests/cassettes/TestLLMHooksIntegration.test_lite_agent_hooks_integration_with_real_llm.yaml

@@ -0,0 +1,87 @@
+interactions:
+- request:
+    body: '{"messages":[{"role":"system","content":"You are Test Assistant. You are
+      a helpful test assistant\nYour personal goal is: Answer questions briefly\n\nTo
+      give my best complete final answer to the task respond using the exact following
+      format:\n\nThought: I now can give a great answer\nFinal Answer: Your final
+      answer must be the great and the most complete as possible, it must be outcome
+      described.\n\nI MUST use these formats, my job depends on it!"},{"role":"user","content":"Say
+      ''Hello World'' and nothing else"}],"model":"gpt-4.1-mini"}'
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate, zstd
+      connection:
+      - keep-alive
+      content-length:
+      - '540'
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.109.1
+      x-stainless-arch:
+      - arm64
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - MacOS
+      x-stainless-package-version:
+      - 1.109.1
+      x-stainless-read-timeout:
+      - '600'
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.13.3
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAAwAAAP//jFLLbtswELzrKxY8W4HlSjasW5Gibfo6FU1fgUCTK4kuxSVIKm4a+N8L
+        So6ltCnQiwDt7Axndvc+AWBKshKYaHkQndXpZXNVv3vz6cWv68/b/XtX0OHuw9v9l92r/uslZ4vI
+        oN0eRXhgXQjqrMagyIywcMgDRtVss86z7aZYPxuAjiTqSGtsSPOLLO2UUelquSrSZZ5m+YnekhLo
+        WQnfEgCA++EbjRqJP1kJy8VDpUPveYOsPDcBMEc6Vhj3XvnATWCLCRRkAprB+8eW+qYNJVyBoQMI
+        bqBRtwgcmhgAuPEHdN/NS2W4hufDXwmvUWuCa3JaznUd1r3nMZzptZ4B3BgKPA5nSHRzQo7nDJoa
+        62jn/6CyWhnl28oh92SiXx/IsgE9JgA3w6z6R/GZddTZUAX6gcNz2XI16rFpRzO0OIGBAtezerZZ
+        PKFXSQxcaT+bNhNctCgn6rQa3ktFMyCZpf7bzVPaY3Jlmv+RnwAh0AaUlXUolXiceGpzGE/4X23n
+        KQ+GmUd3qwRWQaGLm5BY816Pd8X8nQ/YVbUyDTrr1Hhcta22m/Uai3y7W7HkmPwGAAD//wMABY90
+        7msDAAA=
+    headers:
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Wed, 26 Nov 2025 22:52:43 GMT
+      Server:
+      - cloudflare
+      Strict-Transport-Security:
+      - max-age=31536000; includeSubDomains; preload
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - nosniff
+      access-control-expose-headers:
+      - X-Request-ID
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      openai-version:
+      - '2020-10-01'
+      x-openai-proxy-wasm:
+      - v0.1
+    status:
+      code: 200
+      message: OK
+version: 1

lib/crewai/tests/cassettes/hooks/TestLLMHooksIntegration.test_direct_llm_call_hooks_integration.yaml

@@ -0,0 +1,108 @@
+interactions:
+- request:
+    body: '{"messages":[{"role":"user","content":"Say hello"}],"model":"gpt-4o-mini"}'
+    headers:
+      User-Agent:
+      - X-USER-AGENT-XXX
+      accept:
+      - application/json
+      accept-encoding:
+      - ACCEPT-ENCODING-XXX
+      authorization:
+      - AUTHORIZATION-XXX
+      connection:
+      - keep-alive
+      content-length:
+      - '74'
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+      x-stainless-arch:
+      - X-STAINLESS-ARCH-XXX
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - X-STAINLESS-OS-XXX
+      x-stainless-package-version:
+      - 1.83.0
+      x-stainless-read-timeout:
+      - X-STAINLESS-READ-TIMEOUT-XXX
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.13.3
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAAwAAAP//jFKxbtswEN31FVfOViEbRix76RAUcIaiQzu1CASaPMlMSR5Bnhobgf+9
+        oJRYSpMCXTjcu/f43t09FQDCaLEDoY6SlQu2vDXu+5f2HNeeTubuVDl3Wu6/Pn7ufuy/HcQiM+jw
+        gIpfWB8VuWCRDfkRVhElY1Zdbm7Wm7rebG8GwJFGm2ld4HJNpTPelKtqtS6rTbmsn9lHMgqT2MHP
+        AgDgaXizT6/xJHZQLV4qDlOSHYrdtQlARLK5ImRKJrH0LBYTqMgz+sH6Hq2lD7CnR1DSwx2MBDhT
+        D0xanj/NiRHbPsls3vfWzgDpPbHM4QfL98/I5WrSUhciHdJfVNEab9KxiSgT+WwoMQUxoJcC4H4Y
+        Rv8qnwiRXOCG6RcO321HNTFt4C3GxNJO5WW9eEer0cjS2DQbpVBSHVFPzGnusteGZkAxS/zWy3va
+        Y2rju/+RnwClMDDqJkTURr3OO7VFzOf5r7brhAfDImH8bRQ2bDDmLWhsZW/HoxHpnBhd0xrfYQzR
+        jJfThgaVrBTWq20tikvxBwAA//8DAFoGAGtHAwAA
+    headers:
+      CF-RAY:
+      - CF-RAY-XXX
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Wed, 03 Dec 2025 19:06:36 GMT
+      Server:
+      - cloudflare
+      Set-Cookie:
+      - SET-COOKIE-XXX
+      Strict-Transport-Security:
+      - STS-XXX
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - X-CONTENT-TYPE-XXX
+      access-control-expose-headers:
+      - ACCESS-CONTROL-XXX
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      openai-organization:
+      - OPENAI-ORG-XXX
+      openai-processing-ms:
+      - '371'
+      openai-project:
+      - OPENAI-PROJECT-XXX
+      openai-version:
+      - '2020-10-01'
+      x-envoy-upstream-service-time:
+      - '387'
+      x-openai-proxy-wasm:
+      - v0.1
+      x-ratelimit-limit-requests:
+      - X-RATELIMIT-LIMIT-REQUESTS-XXX
+      x-ratelimit-limit-tokens:
+      - X-RATELIMIT-LIMIT-TOKENS-XXX
+      x-ratelimit-remaining-requests:
+      - X-RATELIMIT-REMAINING-REQUESTS-XXX
+      x-ratelimit-remaining-tokens:
+      - X-RATELIMIT-REMAINING-TOKENS-XXX
+      x-ratelimit-reset-requests:
+      - X-RATELIMIT-RESET-REQUESTS-XXX
+      x-ratelimit-reset-tokens:
+      - X-RATELIMIT-RESET-TOKENS-XXX
+      x-request-id:
+      - X-REQUEST-ID-XXX
+    status:
+      code: 200
+      message: OK
+version: 1

lib/crewai/tests/cassettes/hooks/TestLLMHooksIntegration.test_lite_agent_hooks_integration_with_real_llm.yaml

@@ -0,0 +1,115 @@
+interactions:
+- request:
+    body: '{"messages":[{"role":"system","content":"You are Test Assistant. You are
+      a helpful test assistant\nYour personal goal is: Answer questions briefly\n\nTo
+      give my best complete final answer to the task respond using the exact following
+      format:\n\nThought: I now can give a great answer\nFinal Answer: Your final
+      answer must be the great and the most complete as possible, it must be outcome
+      described.\n\nI MUST use these formats, my job depends on it!"},{"role":"user","content":"Say
+      ''Hello World'' and nothing else"}],"model":"gpt-4.1-mini"}'
+    headers:
+      User-Agent:
+      - X-USER-AGENT-XXX
+      accept:
+      - application/json
+      accept-encoding:
+      - ACCEPT-ENCODING-XXX
+      authorization:
+      - AUTHORIZATION-XXX
+      connection:
+      - keep-alive
+      content-length:
+      - '540'
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+      x-stainless-arch:
+      - X-STAINLESS-ARCH-XXX
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - X-STAINLESS-OS-XXX
+      x-stainless-package-version:
+      - 1.83.0
+      x-stainless-read-timeout:
+      - X-STAINLESS-READ-TIMEOUT-XXX
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.13.3
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAAwAAAP//jFLBbtQwEL3nK0Y+b6okZJuSW4VAFIq4rFQkqCLXmSQDjseynRZU7b8j
+        Z7ebFIrEJVLmzXt+b2YeEwBBrahBqEEGNVqdvqFxd331hW+2hD5scyrd5asPxe7j9edPb8UmMvju
+        O6rwxDpTPFqNgdgcYOVQBoyqeXVeVhcX1evtDIzcoo603oa0PMvTkQylRVZs06xM8/JIH5gUelHD
+        1wQA4HH+RqOmxZ+ihmzzVBnRe9mjqE9NAMKxjhUhvScfpAlis4CKTUAze98NPPVDqOEKDD+AkgZ6
+        ukeQ0McAII1/QPfNvCMjNVzOfzW8R60Zbtjpdq3rsJu8jOHMpPUKkMZwkHE4c6LbI7I/ZdDcW8d3
+        /g+q6MiQHxqH0rOJfn1gK2Z0nwDczrOansUX1vFoQxP4B87P5Vlx0BPLjlbo9ggGDlKv6nm1eUGv
+        aTFI0n41baGkGrBdqMtq5NQSr4BklfpvNy9pH5KT6f9HfgGUQhuwbazDltTzxEubw3jC/2o7TXk2
+        LDy6e1LYBEIXN9FiJyd9uCvhf/mAY9OR6dFZR4fj6mxTlFWeqarLzkWyT34DAAD//wMAP95PRmsD
+        AAA=
+    headers:
+      CF-RAY:
+      - CF-RAY-XXX
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Wed, 03 Dec 2025 19:06:36 GMT
+      Server:
+      - cloudflare
+      Set-Cookie:
+      - SET-COOKIE-XXX
+      Strict-Transport-Security:
+      - STS-XXX
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - X-CONTENT-TYPE-XXX
+      access-control-expose-headers:
+      - ACCESS-CONTROL-XXX
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      openai-organization:
+      - OPENAI-ORG-XXX
+      openai-processing-ms:
+      - '620'
+      openai-project:
+      - OPENAI-PROJECT-XXX
+      openai-version:
+      - '2020-10-01'
+      x-envoy-upstream-service-time:
+      - '1891'
+      x-openai-proxy-wasm:
+      - v0.1
+      x-ratelimit-limit-requests:
+      - X-RATELIMIT-LIMIT-REQUESTS-XXX
+      x-ratelimit-limit-tokens:
+      - X-RATELIMIT-LIMIT-TOKENS-XXX
+      x-ratelimit-remaining-requests:
+      - X-RATELIMIT-REMAINING-REQUESTS-XXX
+      x-ratelimit-remaining-tokens:
+      - X-RATELIMIT-REMAINING-TOKENS-XXX
+      x-ratelimit-reset-requests:
+      - X-RATELIMIT-RESET-REQUESTS-XXX
+      x-ratelimit-reset-tokens:
+      - X-RATELIMIT-RESET-TOKENS-XXX
+      x-request-id:
+      - X-REQUEST-ID-XXX
+    status:
+      code: 200
+      message: OK
+version: 1

lib/crewai/tests/cassettes/hooks/TestToolHooksIntegration.test_lite_agent_hooks_integration_with_real_tool.yaml

@@ -0,0 +1,486 @@
+interactions:
+- request:
+    body: '{"messages":[{"role":"system","content":"You are Calculator Assistant.
+      You are a helpful calculator assistant\nYour personal goal is: Help with math
+      calculations\n\nYou ONLY have access to the following tools, and should NEVER
+      make up tools that are not listed here:\n\nTool Name: calculate_sum\nTool Arguments:
+      {''a'': {''description'': None, ''type'': ''int''}, ''b'': {''description'':
+      None, ''type'': ''int''}}\nTool Description: Add two numbers together.\n\nIMPORTANT:
+      Use the following format in your response:\n\n```\nThought: you should always
+      think about what to do\nAction: the action to take, only one name of [calculate_sum],
+      just the name, exactly as it''s written.\nAction Input: the input to the action,
+      just a simple JSON object, enclosed in curly braces, using \" to wrap keys and
+      values.\nObservation: the result of the action\n```\n\nOnce all necessary information
+      is gathered, return the following format:\n\n```\nThought: I now know the final
+      answer\nFinal Answer: the final answer to the original input question\n```"},{"role":"user","content":"What
+      is 5 + 3? Use the calculate_sum tool."}],"model":"gpt-4.1-mini"}'
+    headers:
+      User-Agent:
+      - X-USER-AGENT-XXX
+      accept:
+      - application/json
+      accept-encoding:
+      - ACCEPT-ENCODING-XXX
+      authorization:
+      - AUTHORIZATION-XXX
+      connection:
+      - keep-alive
+      content-length:
+      - '1119'
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+      x-stainless-arch:
+      - X-STAINLESS-ARCH-XXX
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - X-STAINLESS-OS-XXX
+      x-stainless-package-version:
+      - 1.83.0
+      x-stainless-read-timeout:
+      - X-STAINLESS-READ-TIMEOUT-XXX
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.13.3
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAAwAAAP//jFPLbtswELzrKxY8W0Zsy3ajWxv0EbSXFm0CtAokmlxLTCiSIFdFWsP/
+        XlB+SO4D6IUQdnZGs7PkLgFgSrIcmGg4idbp9EY9hbvZsrn7cP/l0/uv77JXzx9fv31UN2/u3U82
+        iQy7eURBJ9ZU2NZpJGXNARYeOWFUna1X2frFfLVa90BrJepIqx2l2XSWtsqodH41X6ZXWTrLjvTG
+        KoGB5fAtAQDY9Wc0aiQ+sxyuJqdKiyHwGll+bgJg3upYYTwEFYgbYpMBFNYQmt57VVWF+dzYrm4o
+        h1sIje20hC4gUIMguBad5oRl6FogazWQBS4lLIEbCYtpYV6KOHV+2Xsqw61xHeWwKxgvWA7LCRRs
+        E78W+8JUVTX25XHbBR7DMZ3WI4AbY4lHvT6RhyOyP2egbe283YTfqGyrjApN6ZEHa+K8gaxjPbpP
+        AB76rLuL+JjztnVUkn3C/nfzRXbQY8OOBzQ7LoKRJa5HrPWJdaFXSiSudBhtiwkuGpQDdVgt76Sy
+        IyAZTf2nm79pHyZXpv4f+QEQAh2hLJ1HqcTlxEObx/gE/tV2Trk3zAL670pgSQp93ITELe/04V6y
+        8CMQtuVWmRq98+pwObeuvF6vVrjMrjdzluyTXwAAAP//AwAHWkpkqwMAAA==
+    headers:
+      CF-RAY:
+      - CF-RAY-XXX
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Wed, 03 Dec 2025 17:24:28 GMT
+      Server:
+      - cloudflare
+      Set-Cookie:
+      - SET-COOKIE-XXX
+      Strict-Transport-Security:
+      - STS-XXX
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - X-CONTENT-TYPE-XXX
+      access-control-expose-headers:
+      - ACCESS-CONTROL-XXX
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      openai-organization:
+      - OPENAI-ORG-XXX
+      openai-processing-ms:
+      - '681'
+      openai-project:
+      - OPENAI-PROJECT-XXX
+      openai-version:
+      - '2020-10-01'
+      x-envoy-upstream-service-time:
+      - '871'
+      x-openai-proxy-wasm:
+      - v0.1
+      x-ratelimit-limit-requests:
+      - X-RATELIMIT-LIMIT-REQUESTS-XXX
+      x-ratelimit-limit-tokens:
+      - X-RATELIMIT-LIMIT-TOKENS-XXX
+      x-ratelimit-remaining-requests:
+      - X-RATELIMIT-REMAINING-REQUESTS-XXX
+      x-ratelimit-remaining-tokens:
+      - X-RATELIMIT-REMAINING-TOKENS-XXX
+      x-ratelimit-reset-requests:
+      - X-RATELIMIT-RESET-REQUESTS-XXX
+      x-ratelimit-reset-tokens:
+      - X-RATELIMIT-RESET-TOKENS-XXX
+      x-request-id:
+      - X-REQUEST-ID-XXX
+    status:
+      code: 200
+      message: OK
+- request:
+    body: '{"messages":[{"role":"system","content":"You are Calculator Assistant.
+      You are a helpful calculator assistant\nYour personal goal is: Help with math
+      calculations\n\nYou ONLY have access to the following tools, and should NEVER
+      make up tools that are not listed here:\n\nTool Name: calculate_sum\nTool Arguments:
+      {''a'': {''description'': None, ''type'': ''int''}, ''b'': {''description'':
+      None, ''type'': ''int''}}\nTool Description: Add two numbers together.\n\nIMPORTANT:
+      Use the following format in your response:\n\n```\nThought: you should always
+      think about what to do\nAction: the action to take, only one name of [calculate_sum],
+      just the name, exactly as it''s written.\nAction Input: the input to the action,
+      just a simple JSON object, enclosed in curly braces, using \" to wrap keys and
+      values.\nObservation: the result of the action\n```\n\nOnce all necessary information
+      is gathered, return the following format:\n\n```\nThought: I now know the final
+      answer\nFinal Answer: the final answer to the original input question\n```"},{"role":"user","content":"What
+      is 5 + 3? Use the calculate_sum tool."},{"role":"assistant","content":"```\nThought:
+      I should use the calculate_sum tool to add 5 and 3.\nAction: calculate_sum\nAction
+      Input: {\"a\": 5, \"b\": 3}\n```\nObservation: 8"}],"model":"gpt-4.1-mini"}'
+    headers:
+      User-Agent:
+      - X-USER-AGENT-XXX
+      accept:
+      - application/json
+      accept-encoding:
+      - ACCEPT-ENCODING-XXX
+      authorization:
+      - AUTHORIZATION-XXX
+      connection:
+      - keep-alive
+      content-length:
+      - '1298'
+      content-type:
+      - application/json
+      cookie:
+      - COOKIE-XXX
+      host:
+      - api.openai.com
+      x-stainless-arch:
+      - X-STAINLESS-ARCH-XXX
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - X-STAINLESS-OS-XXX
+      x-stainless-package-version:
+      - 1.83.0
+      x-stainless-read-timeout:
+      - X-STAINLESS-READ-TIMEOUT-XXX
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.13.3
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAAwAAAP//jFJNb5wwEL3zKyyfl2ghBAi3KtJKTS9VlX5I3QiMGcBZY7v20DSK9r9X
+        hs1C2lTqxdLMm/c8b2aeA0KoaGhBKO8Z8sHI8EYc3Ff7pX663Y32w82Qffv46XNUR+Yu+7GjG8/Q
+        9QNwfGFdcD0YCSi0mmFugSF41ShLkyyP0zSfgEE3ID2tMxgmF1E4CCXCeBtfhdskjJITvdeCg6MF
+        +R4QQsjz9PpGVQO/aEG2m5fMAM6xDmhxLiKEWi19hjLnhEOmkG4WkGuFoKbeq6raq7tej12PBXlP
+        lH4kB/9gD6QViknClHsEu1e7KXo3RQXJ96qqqrWqhXZ0zFtTo5QrgCmlkfnRTH7uT8jx7EDqzlhd
+        uz+otBVKuL60wJxWvluH2tAJPQaE3E+TGl+Zp8bqwWCJ+gDTd3F+OevRZUMLGuUnEDUyueQvt9Hm
+        Db2yAWRCutWsKWe8h2ahLothYyP0CghWrv/u5i3t2blQ3f/ILwDnYBCa0lhoBH/teCmz4A/4X2Xn
+        KU8NUwf2p+BQogDrN9FAy0Y5XxV1Tw5hKFuhOrDGivm0WlNeZ2kKV8l1HdPgGPwGAAD//wMAJksH
+        jGkDAAA=
+    headers:
+      CF-RAY:
+      - CF-RAY-XXX
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Wed, 03 Dec 2025 17:24:29 GMT
+      Server:
+      - cloudflare
+      Strict-Transport-Security:
+      - STS-XXX
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - X-CONTENT-TYPE-XXX
+      access-control-expose-headers:
+      - ACCESS-CONTROL-XXX
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      openai-organization:
+      - OPENAI-ORG-XXX
+      openai-processing-ms:
+      - '427'
+      openai-project:
+      - OPENAI-PROJECT-XXX
+      openai-version:
+      - '2020-10-01'
+      x-envoy-upstream-service-time:
+      - '442'
+      x-openai-proxy-wasm:
+      - v0.1
+      x-ratelimit-limit-requests:
+      - X-RATELIMIT-LIMIT-REQUESTS-XXX
+      x-ratelimit-limit-tokens:
+      - X-RATELIMIT-LIMIT-TOKENS-XXX
+      x-ratelimit-remaining-requests:
+      - X-RATELIMIT-REMAINING-REQUESTS-XXX
+      x-ratelimit-remaining-tokens:
+      - X-RATELIMIT-REMAINING-TOKENS-XXX
+      x-ratelimit-reset-requests:
+      - X-RATELIMIT-RESET-REQUESTS-XXX
+      x-ratelimit-reset-tokens:
+      - X-RATELIMIT-RESET-TOKENS-XXX
+      x-request-id:
+      - X-REQUEST-ID-XXX
+    status:
+      code: 200
+      message: OK
+- request:
+    body: '{"messages":[{"role":"system","content":"You are Calculator Assistant.
+      You are a helpful calculator assistant\nYour personal goal is: Help with math
+      calculations\n\nYou ONLY have access to the following tools, and should NEVER
+      make up tools that are not listed here:\n\nTool Name: calculate_sum\nTool Arguments:
+      {''a'': {''description'': None, ''type'': ''int''}, ''b'': {''description'':
+      None, ''type'': ''int''}}\nTool Description: Add two numbers together.\n\nIMPORTANT:
+      Use the following format in your response:\n\n```\nThought: you should always
+      think about what to do\nAction: the action to take, only one name of [calculate_sum],
+      just the name, exactly as it''s written.\nAction Input: the input to the action,
+      just a simple JSON object, enclosed in curly braces, using \" to wrap keys and
+      values.\nObservation: the result of the action\n```\n\nOnce all necessary information
+      is gathered, return the following format:\n\n```\nThought: I now know the final
+      answer\nFinal Answer: the final answer to the original input question\n```"},{"role":"user","content":"What
+      is 5 + 3? Use the calculate_sum tool."}],"model":"gpt-4.1-mini"}'
+    headers:
+      User-Agent:
+      - X-USER-AGENT-XXX
+      accept:
+      - application/json
+      accept-encoding:
+      - ACCEPT-ENCODING-XXX
+      authorization:
+      - AUTHORIZATION-XXX
+      connection:
+      - keep-alive
+      content-length:
+      - '1119'
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+      x-stainless-arch:
+      - X-STAINLESS-ARCH-XXX
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - X-STAINLESS-OS-XXX
+      x-stainless-package-version:
+      - 1.83.0
+      x-stainless-read-timeout:
+      - X-STAINLESS-READ-TIMEOUT-XXX
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.13.3
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAAwAAAP//jJLLbtswEEX3+ooB15bhh2zH2iXtJovuYqAPBRJNjiU6FEmQo7ap4X8v
+        JDmW3KZAN1rMmXs1c4enCIApyVJgouIkaqfjD6r+fFf9+KJ2u6/qeO92H1/mrw+ffj25h+OWTVqF
+        3R9R0JtqKmztNJKypsfCIydsXeebdbK5286Wsw7UVqJuZaWjOJnO41oZFS9mi1U8S+J5cpFXVgkM
+        LIVvEQDAqfu2gxqJP1kKnVlXqTEEXiJLr00AzFvdVhgPQQXihthkgMIaQtPNXhRFZp4q25QVpfAI
+        BlECWeBSwgq4kbCEJihTAlUIgmvRaE6Yh6YGslZPM3Mv2q3TW/hWhkfjGkrhlDGesXQ1ydg+Y+ny
+        nJmiKMZDeTw0gbfJmEbrEeDGWOKtWRfH84WcrwFoWzpv9+EPKTsoo0KVe+TBmnbZQNaxjp4jgOcu
+        6OYmO+a8rR3lZF+w+91imfR+bDjwQJebCyRLXI9Um/nkHb9cInGlw+hUTHBRoRykw115I5UdgWi0
+        9d/TvOfdb65M+T/2AxACHaHMnUepxO3GQ5vH9v3/q+2acjcwC+i/K4E5KfTtJSQeeKP7R8nCayCs
+        84MyJXrnVf8yDy7fbtZrXCXb/YJF5+g3AAAA//8DANrSB6yoAwAA
+    headers:
+      CF-RAY:
+      - CF-RAY-XXX
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Wed, 03 Dec 2025 19:10:33 GMT
+      Server:
+      - cloudflare
+      Set-Cookie:
+      - SET-COOKIE-XXX
+      Strict-Transport-Security:
+      - STS-XXX
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - X-CONTENT-TYPE-XXX
+      access-control-expose-headers:
+      - ACCESS-CONTROL-XXX
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      openai-organization:
+      - OPENAI-ORG-XXX
+      openai-processing-ms:
+      - '2329'
+      openai-project:
+      - OPENAI-PROJECT-XXX
+      openai-version:
+      - '2020-10-01'
+      x-envoy-upstream-service-time:
+      - '2349'
+      x-openai-proxy-wasm:
+      - v0.1
+      x-ratelimit-limit-requests:
+      - X-RATELIMIT-LIMIT-REQUESTS-XXX
+      x-ratelimit-limit-tokens:
+      - X-RATELIMIT-LIMIT-TOKENS-XXX
+      x-ratelimit-remaining-requests:
+      - X-RATELIMIT-REMAINING-REQUESTS-XXX
+      x-ratelimit-remaining-tokens:
+      - X-RATELIMIT-REMAINING-TOKENS-XXX
+      x-ratelimit-reset-requests:
+      - X-RATELIMIT-RESET-REQUESTS-XXX
+      x-ratelimit-reset-tokens:
+      - X-RATELIMIT-RESET-TOKENS-XXX
+      x-request-id:
+      - X-REQUEST-ID-XXX
+    status:
+      code: 200
+      message: OK
+- request:
+    body: '{"messages":[{"role":"system","content":"You are Calculator Assistant.
+      You are a helpful calculator assistant\nYour personal goal is: Help with math
+      calculations\n\nYou ONLY have access to the following tools, and should NEVER
+      make up tools that are not listed here:\n\nTool Name: calculate_sum\nTool Arguments:
+      {''a'': {''description'': None, ''type'': ''int''}, ''b'': {''description'':
+      None, ''type'': ''int''}}\nTool Description: Add two numbers together.\n\nIMPORTANT:
+      Use the following format in your response:\n\n```\nThought: you should always
+      think about what to do\nAction: the action to take, only one name of [calculate_sum],
+      just the name, exactly as it''s written.\nAction Input: the input to the action,
+      just a simple JSON object, enclosed in curly braces, using \" to wrap keys and
+      values.\nObservation: the result of the action\n```\n\nOnce all necessary information
+      is gathered, return the following format:\n\n```\nThought: I now know the final
+      answer\nFinal Answer: the final answer to the original input question\n```"},{"role":"user","content":"What
+      is 5 + 3? Use the calculate_sum tool."},{"role":"assistant","content":"```\nThought:
+      I need to add 5 and 3 using the calculate_sum tool.\nAction: calculate_sum\nAction
+      Input: {\"a\":5,\"b\":3}\n```\nObservation: 8"}],"model":"gpt-4.1-mini"}'
+    headers:
+      User-Agent:
+      - X-USER-AGENT-XXX
+      accept:
+      - application/json
+      accept-encoding:
+      - ACCEPT-ENCODING-XXX
+      authorization:
+      - AUTHORIZATION-XXX
+      connection:
+      - keep-alive
+      content-length:
+      - '1295'
+      content-type:
+      - application/json
+      cookie:
+      - COOKIE-XXX
+      host:
+      - api.openai.com
+      x-stainless-arch:
+      - X-STAINLESS-ARCH-XXX
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - X-STAINLESS-OS-XXX
+      x-stainless-package-version:
+      - 1.83.0
+      x-stainless-read-timeout:
+      - X-STAINLESS-READ-TIMEOUT-XXX
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.13.3
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAAwAAAP//jJJNb9QwEIbv+RWWz5tqN81+5QaVQHDi0AOIrRKvPUncOmNjTyhQ7X9H
+        zn4kS0HiYsl+5h3POzMvCWNcK14wLltBsnMmvdPd57f+yzKovc/cR/z23ohfn+69yJX1fBYVdv8I
+        ks6qG2k7Z4C0xSOWHgRBzLpYr/L1Zju/vR1AZxWYKGscpfnNIu006jSbZ8t0nqeL/CRvrZYQeMG+
+        Jowx9jKcsVBU8IMXbD47v3QQgmiAF5cgxri3Jr5wEYIOJJD4bITSIgEOtVdVtcP71vZNSwX7wNA+
+        s6d4UAus1igMExiewe/w3XB7M9wKttlhVVXTrB7qPohoDXtjJkAgWhKxNYOfhxM5XBwY2zhv9+EP
+        Ka816tCWHkSwGKsNZB0f6CFh7GHoVH9lnjtvO0cl2ScYvss2p07xcUIjXWxOkCwJM1Ftz+AqX6mA
+        hDZh0msuhWxBjdJxMKJX2k5AMnH9upq/5T4619j8T/oRSAmOQJXOg9Ly2vEY5iEu8L/CLl0eCuYB
+        /HctoSQNPk5CQS16c9wqHn4Ggq6sNTbgndfH1apduV2vVrDMt/uMJ4fkNwAAAP//AwCRC7shaQMA
+        AA==
+    headers:
+      CF-RAY:
+      - CF-RAY-XXX
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Wed, 03 Dec 2025 19:10:35 GMT
+      Server:
+      - cloudflare
+      Strict-Transport-Security:
+      - STS-XXX
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - X-CONTENT-TYPE-XXX
+      access-control-expose-headers:
+      - ACCESS-CONTROL-XXX
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      openai-organization:
+      - OPENAI-ORG-XXX
+      openai-processing-ms:
+      - '1647'
+      openai-project:
+      - OPENAI-PROJECT-XXX
+      openai-version:
+      - '2020-10-01'
+      x-envoy-upstream-service-time:
+      - '1694'
+      x-openai-proxy-wasm:
+      - v0.1
+      x-ratelimit-limit-requests:
+      - X-RATELIMIT-LIMIT-REQUESTS-XXX
+      x-ratelimit-limit-tokens:
+      - X-RATELIMIT-LIMIT-TOKENS-XXX
+      x-ratelimit-remaining-requests:
+      - X-RATELIMIT-REMAINING-REQUESTS-XXX
+      x-ratelimit-remaining-tokens:
+      - X-RATELIMIT-REMAINING-TOKENS-XXX
+      x-ratelimit-reset-requests:
+      - X-RATELIMIT-RESET-REQUESTS-XXX
+      x-ratelimit-reset-tokens:
+      - X-RATELIMIT-RESET-TOKENS-XXX
+      x-request-id:
+      - X-REQUEST-ID-XXX
+    status:
+      code: 200
+      message: OK
+version: 1

lib/crewai/tests/hooks/test_llm_hooks.py

@@ -309,3 +309,188 @@ class TestLLMHooksIntegration:
         clear_all_llm_call_hooks()
         hooks = get_before_llm_call_hooks()
         assert len(hooks) == 0
+
+    @pytest.mark.vcr()
+    def test_lite_agent_hooks_integration_with_real_llm(self):
+        """Test that LiteAgent executes before/after LLM call hooks and prints messages correctly."""
+        import os
+        from crewai.lite_agent import LiteAgent
+
+        # Skip if no API key available
+        if not os.environ.get("OPENAI_API_KEY"):
+            pytest.skip("OPENAI_API_KEY not set - skipping real LLM test")
+
+        # Track hook invocations
+        hook_calls = {"before": [], "after": []}
+
+        def before_llm_call_hook(context: LLMCallHookContext) -> bool:
+            """Log and verify before hook execution."""
+            print(f"\n[BEFORE HOOK] Agent: {context.agent.role if context.agent else 'None'}")
+            print(f"[BEFORE HOOK] Iterations: {context.iterations}")
+            print(f"[BEFORE HOOK] Message count: {len(context.messages)}")
+            print(f"[BEFORE HOOK] Messages: {context.messages}")
+
+            # Track the call
+            hook_calls["before"].append({
+                "iterations": context.iterations,
+                "message_count": len(context.messages),
+                "has_task": context.task is not None,
+                "has_crew": context.crew is not None,
+            })
+
+            return True  # Allow execution
+
+        def after_llm_call_hook(context: LLMCallHookContext) -> str | None:
+            """Log and verify after hook execution."""
+            print(f"\n[AFTER HOOK] Agent: {context.agent.role if context.agent else 'None'}")
+            print(f"[AFTER HOOK] Iterations: {context.iterations}")
+            print(f"[AFTER HOOK] Response: {context.response[:100] if context.response else 'None'}...")
+            print(f"[AFTER HOOK] Final message count: {len(context.messages)}")
+
+            # Track the call
+            hook_calls["after"].append({
+                "iterations": context.iterations,
+                "has_response": context.response is not None,
+                "response_length": len(context.response) if context.response else 0,
+            })
+
+            # Optionally modify response
+            if context.response:
+                return f"[HOOKED] {context.response}"
+            return None
+
+        # Register hooks
+        register_before_llm_call_hook(before_llm_call_hook)
+        register_after_llm_call_hook(after_llm_call_hook)
+
+        try:
+            # Create LiteAgent
+            lite_agent = LiteAgent(
+                role="Test Assistant",
+                goal="Answer questions briefly",
+                backstory="You are a helpful test assistant",
+                verbose=True,
+            )
+
+            # Verify hooks are loaded
+            assert len(lite_agent.before_llm_call_hooks) > 0, "Before hooks not loaded"
+            assert len(lite_agent.after_llm_call_hooks) > 0, "After hooks not loaded"
+
+            # Execute with a simple prompt
+            result = lite_agent.kickoff("Say 'Hello World' and nothing else")
+
+
+            # Verify hooks were called
+            assert len(hook_calls["before"]) > 0, "Before hook was never called"
+            assert len(hook_calls["after"]) > 0, "After hook was never called"
+
+            # Verify context had correct attributes for LiteAgent (used in flows)
+            # LiteAgent doesn't have task/crew context, unlike agents in CrewBase
+            before_call = hook_calls["before"][0]
+            assert before_call["has_task"] is False, "Task should be None for LiteAgent in flows"
+            assert before_call["has_crew"] is False, "Crew should be None for LiteAgent in flows"
+            assert before_call["message_count"] > 0, "Should have messages"
+
+            # Verify after hook received response
+            after_call = hook_calls["after"][0]
+            assert after_call["has_response"] is True, "After hook should have response"
+            assert after_call["response_length"] > 0, "Response should not be empty"
+
+            # Verify response was modified by after hook
+            # Note: The hook modifies the raw LLM response, but LiteAgent then parses it
+            # to extract the "Final Answer" portion. We check the messages to see the modification.
+            assert len(result.messages) > 2, "Should have assistant message in messages"
+            last_message = result.messages[-1]
+            assert last_message["role"] == "assistant", "Last message should be from assistant"
+            assert "[HOOKED]" in last_message["content"], "Hook should have modified the assistant message"
+
+
+        finally:
+            # Clean up hooks
+            unregister_before_llm_call_hook(before_llm_call_hook)
+            unregister_after_llm_call_hook(after_llm_call_hook)
+
+    @pytest.mark.vcr()
+    def test_direct_llm_call_hooks_integration(self):
+        """Test that hooks work for direct llm.call() without agents."""
+        import os
+        from crewai.llm import LLM
+
+        # Skip if no API key available
+        if not os.environ.get("OPENAI_API_KEY"):
+            pytest.skip("OPENAI_API_KEY not set - skipping real LLM test")
+
+        # Track hook invocations
+        hook_calls = {"before": [], "after": []}
+
+        def before_hook(context: LLMCallHookContext) -> bool:
+            """Log and verify before hook execution."""
+            print(f"\n[BEFORE HOOK] Agent: {context.agent}")
+            print(f"[BEFORE HOOK] Task: {context.task}")
+            print(f"[BEFORE HOOK] Crew: {context.crew}")
+            print(f"[BEFORE HOOK] LLM: {context.llm}")
+            print(f"[BEFORE HOOK] Iterations: {context.iterations}")
+            print(f"[BEFORE HOOK] Message count: {len(context.messages)}")
+
+            # Track the call
+            hook_calls["before"].append({
+                "agent": context.agent,
+                "task": context.task,
+                "crew": context.crew,
+                "llm": context.llm is not None,
+                "message_count": len(context.messages),
+            })
+
+            return True  # Allow execution
+
+        def after_hook(context: LLMCallHookContext) -> str | None:
+            """Log and verify after hook execution."""
+            print(f"\n[AFTER HOOK] Agent: {context.agent}")
+            print(f"[AFTER HOOK] Response: {context.response[:100] if context.response else 'None'}...")
+
+            # Track the call
+            hook_calls["after"].append({
+                "has_response": context.response is not None,
+                "response_length": len(context.response) if context.response else 0,
+            })
+
+            # Modify response
+            if context.response:
+                return f"[HOOKED] {context.response}"
+            return None
+
+        # Register hooks
+        register_before_llm_call_hook(before_hook)
+        register_after_llm_call_hook(after_hook)
+
+        try:
+            # Create LLM and make direct call
+            llm = LLM(model="gpt-4o-mini")
+            result = llm.call([{"role": "user", "content": "Say hello"}])
+
+            print(f"\n[TEST] Final result: {result}")
+
+            # Verify hooks were called
+            assert len(hook_calls["before"]) > 0, "Before hook was never called"
+            assert len(hook_calls["after"]) > 0, "After hook was never called"
+
+            # Verify context had correct attributes for direct LLM calls
+            before_call = hook_calls["before"][0]
+            assert before_call["agent"] is None, "Agent should be None for direct LLM calls"
+            assert before_call["task"] is None, "Task should be None for direct LLM calls"
+            assert before_call["crew"] is None, "Crew should be None for direct LLM calls"
+            assert before_call["llm"] is True, "LLM should be present"
+            assert before_call["message_count"] > 0, "Should have messages"
+
+            # Verify after hook received response
+            after_call = hook_calls["after"][0]
+            assert after_call["has_response"] is True, "After hook should have response"
+            assert after_call["response_length"] > 0, "Response should not be empty"
+
+            # Verify response was modified by after hook
+            assert "[HOOKED]" in result, "Response should be modified by after hook"
+
+        finally:
+            # Clean up hooks
+            unregister_before_llm_call_hook(before_hook)
+            unregister_after_llm_call_hook(after_hook)

lib/crewai/tests/hooks/test_tool_hooks.py

@@ -496,3 +496,97 @@ class TestToolHooksIntegration:
         clear_all_tool_call_hooks()
         hooks = get_before_tool_call_hooks()
         assert len(hooks) == 0
+
+    @pytest.mark.vcr()
+    def test_lite_agent_hooks_integration_with_real_tool(self):
+        """Test that LiteAgent executes before/after tool call hooks with real tool calls."""
+        import os
+        from crewai.lite_agent import LiteAgent
+        from crewai.tools import tool
+
+        # Skip if no API key available
+        if not os.environ.get("OPENAI_API_KEY"):
+            pytest.skip("OPENAI_API_KEY not set - skipping real tool test")
+
+        # Track hook invocations
+        hook_calls = {"before": [], "after": []}
+
+        # Create a simple test tool
+        @tool("calculate_sum")
+        def calculate_sum(a: int, b: int) -> int:
+            """Add two numbers together."""
+            return a + b
+
+        def before_tool_call_hook(context: ToolCallHookContext) -> bool:
+            """Log and verify before hook execution."""
+            print(f"\n[BEFORE HOOK] Tool: {context.tool_name}")
+            print(f"[BEFORE HOOK] Tool input: {context.tool_input}")
+            print(f"[BEFORE HOOK] Agent: {context.agent.role if context.agent else 'None'}")
+            print(f"[BEFORE HOOK] Task: {context.task}")
+            print(f"[BEFORE HOOK] Crew: {context.crew}")
+
+            # Track the call
+            hook_calls["before"].append({
+                "tool_name": context.tool_name,
+                "tool_input": context.tool_input,
+                "has_agent": context.agent is not None,
+                "has_task": context.task is not None,
+                "has_crew": context.crew is not None,
+            })
+
+            return True  # Allow execution
+
+        def after_tool_call_hook(context: ToolCallHookContext) -> str | None:
+            """Log and verify after hook execution."""
+            print(f"\n[AFTER HOOK] Tool: {context.tool_name}")
+            print(f"[AFTER HOOK] Tool result: {context.tool_result}")
+            print(f"[AFTER HOOK] Agent: {context.agent.role if context.agent else 'None'}")
+
+            # Track the call
+            hook_calls["after"].append({
+                "tool_name": context.tool_name,
+                "tool_result": context.tool_result,
+                "has_result": context.tool_result is not None,
+            })
+
+            return None  # Don't modify result
+
+        # Register hooks
+        register_before_tool_call_hook(before_tool_call_hook)
+        register_after_tool_call_hook(after_tool_call_hook)
+
+        try:
+            # Create LiteAgent with the tool
+            lite_agent = LiteAgent(
+                role="Calculator Assistant",
+                goal="Help with math calculations",
+                backstory="You are a helpful calculator assistant",
+                tools=[calculate_sum],
+                verbose=True,
+            )
+
+            # Execute with a prompt that should trigger tool usage
+            result = lite_agent.kickoff("What is 5 + 3? Use the calculate_sum tool.")
+
+            # Verify hooks were called
+            assert len(hook_calls["before"]) > 0, "Before hook was never called"
+            assert len(hook_calls["after"]) > 0, "After hook was never called"
+
+            # Verify context had correct attributes for LiteAgent (used in flows)
+            # LiteAgent doesn't have task/crew context, unlike agents in CrewBase
+            before_call = hook_calls["before"][0]
+            assert before_call["tool_name"] == "calculate_sum", "Tool name should be 'calculate_sum'"
+            assert "a" in before_call["tool_input"], "Tool input should have 'a' parameter"
+            assert "b" in before_call["tool_input"], "Tool input should have 'b' parameter"
+
+            # Verify after hook received result
+            after_call = hook_calls["after"][0]
+            assert after_call["has_result"] is True, "After hook should have tool result"
+            assert after_call["tool_name"] == "calculate_sum", "Tool name should match"
+            # The result should contain the sum (8)
+            assert "8" in str(after_call["tool_result"]), "Tool result should contain the sum"
+
+        finally:
+            # Clean up hooks
+            unregister_before_tool_call_hook(before_tool_call_hook)
+            unregister_after_tool_call_hook(after_tool_call_hook)