Hooks

Learn how to use tool hooks to modify the behavior of a tool.

Tool Hooks

You can use tool hooks to perform validation, logging, or any other logic before or after a tool is called.

A tool hook is a function that takes a function name, function call, and arguments. Optionally, you can access the Agent or Team object as well. Inside the tool hook, you have to call the function call and return the result.

It is important to use exact parameter names when defining a tool hook. agent, team, run_context, function_name, function_call, and arguments are available parameters.

For example:

def logger_hook(
    function_name: str, function_call: Callable, arguments: Dict[str, Any]
):
    """Log the duration of the function call"""
    start_time = time.time()

    # Call the function
    result = function_call(**arguments)

    end_time = time.time()
    duration = end_time - start_time

    logger.info(f"Function {function_name} took {duration:.2f} seconds to execute")

    # Return the result
    return result

or

def confirmation_hook(
    function_name: str, function_call: Callable, arguments: Dict[str, Any]
):
    """Confirm the function call"""
    if function_name != "get_top_hackernews_stories":
        raise ValueError("This tool is not allowed to be called")
    return function_call(**arguments)

You can assign tool hooks on agents and teams. The tool hooks will be applied to all tool calls made by the agent or team.

For example:

agent = Agent(
    model=OpenAIChat(id="gpt-5-mini"),
    tools=[DuckDuckGoTools()],
    tool_hooks=[logger_hook],
)

You can also get access to the RunContext object in the tool hook. Inside the run context, you will find the session state, dependencies, and metadata.

from agno.run import RunContext

def grab_customer_profile_hook(
    run_context: RunContext, function_name: str, function_call: Callable, arguments: Dict[str, Any]
):
    if not run_context.session_state:
        run_context.session_state = {}

    cust_id = arguments.get("customer")
    if cust_id not in run_context.session_state["customer_profiles"]:
        raise ValueError(f"Customer profile for {cust_id} not found")
    customer_profile = run_context.session_state["customer_profiles"][cust_id]

    # Replace the customer with the customer_profile for the function call
    arguments["customer"] = json.dumps(customer_profile)
    # Call the function with the updated arguments
    result = function_call(**arguments)

    return result

Multiple Tool Hooks

You can also assign multiple tool hooks at once. They will be applied in the order they are assigned.

agent = Agent(
    model=OpenAIChat(id="gpt-5-mini"),
    tools=[DuckDuckGoTools()],
    tool_hooks=[logger_hook, confirmation_hook],  # The logger_hook will run on the outer layer, and the confirmation_hook will run on the inner layer
)

You can also assign tool hooks to specific custom tools.

@tool(tool_hooks=[logger_hook, confirmation_hook])
def get_top_hackernews_stories(num_stories: int) -> Iterator[str]:
    """Fetch top stories from Hacker News.

    Args:
        num_stories (int): Number of stories to retrieve
    """
    # Fetch top story IDs
    response = httpx.get("https://hacker-news.firebaseio.com/v0/topstories.json")
    story_ids = response.json()

    # Yield story details
    final_stories = []
    for story_id in story_ids[:num_stories]:
        story_response = httpx.get(
            f"https://hacker-news.firebaseio.com/v0/item/{story_id}.json"
        )
        story = story_response.json()
        if "text" in story:
            story.pop("text", None)
        final_stories.append(story)

    return json.dumps(final_stories)

agent = Agent(
    model=OpenAIChat(id="gpt-5-mini"),
    tools=[get_top_hackernews_stories],
)

Pre and Post Hooks

Pre and post hooks let’s you modify what happens before and after a tool is called. It is an alternative to tool hooks.

Set the pre_hook in the @tool decorator to run a function before the tool call.

Set the post_hook in the @tool decorator to run a function after the tool call.

Here’s a demo example of using a pre_hook, post_hook along with Agent Context.

pre_and_post_hooks.py

import json
from typing import Iterator

import httpx
from agno.agent import Agent
from agno.tools import FunctionCall, tool


def pre_hook(fc: FunctionCall):
    print(f"Pre-hook: {fc.function.name}")
    print(f"Arguments: {fc.arguments}")
    print(f"Result: {fc.result}")


def post_hook(fc: FunctionCall):
    print(f"Post-hook: {fc.function.name}")
    print(f"Arguments: {fc.arguments}")
    print(f"Result: {fc.result}")


@tool(pre_hook=pre_hook, post_hook=post_hook)
def get_top_hackernews_stories(agent: Agent) -> Iterator[str]:
    num_stories = agent.context.get("num_stories", 5) if agent.context else 5

    # Fetch top story IDs
    response = httpx.get("https://hacker-news.firebaseio.com/v0/topstories.json")
    story_ids = response.json()

    # Yield story details
    for story_id in story_ids[:num_stories]:
        story_response = httpx.get(
            f"https://hacker-news.firebaseio.com/v0/item/{story_id}.json"
        )
        story = story_response.json()
        if "text" in story:
            story.pop("text", None)
        yield json.dumps(story)


agent = Agent(
    dependencies={
        "num_stories": 2,
    },
    tools=[get_top_hackernews_stories],
    markdown=True,
)
agent.print_response("What are the top hackernews stories?", stream=True)

Example: Human in the loop using tool hooks

This example shows how to:

• Add hooks to tools for user confirmation
• Handle user input during tool execution
• Gracefully cancel operations based on user choice

1. Create the example

   hitl.py

   """🤝 Human-in-the-Loop: Adding User Confirmation to Tool Calls
   
   This example shows how to implement human-in-the-loop functionality in your Agno tools.
   It shows how to:
   - Add tool hooks to tools for user confirmation
   - Handle user input during tool execution
   - Gracefully cancel operations based on user choice
   
   Some practical applications:
   - Confirming sensitive operations before execution
   - Reviewing API calls before they're made
   - Validating data transformations
   - Approving automated actions in critical systems
   
   Run `pip install openai httpx rich agno` to install dependencies.
   """
   
   import json
   from typing import Any, Callable, Dict, Iterator
   
   import httpx
   from agno.agent import Agent
   from agno.exceptions import StopAgentRun
   from agno.models.openai import OpenAIChat
   from agno.tools import FunctionCall, tool
   from rich.console import Console
   from rich.pretty import pprint
   from rich.prompt import Prompt
   
   # This is the console instance used by the print_response method
   # We can use this to stop and restart the live display and ask for user confirmation
   console = Console()
   
   
   def confirmation_hook(
       function_name: str, function_call: Callable, arguments: Dict[str, Any]
   ):
       # Get the live display instance from the console
       live = console._live
   
       # Stop the live display temporarily so we can ask for user confirmation
       live.stop()  # type: ignore
   
       # Ask for confirmation
       console.print(f"\nAbout to run [bold blue]{fc.function.name}[/]")
       message = (
           Prompt.ask("Do you want to continue?", choices=["y", "n"], default="y")
           .strip()
           .lower()
       )
   
       # Restart the live display
       live.start()  # type: ignore
   
       # If the user does not want to continue, raise a StopExecution exception
       if message != "y":
           raise StopAgentRun(
               "Tool call cancelled by user",
               agent_message="Stopping execution as permission was not granted.",
           )
   
       # Call the function
       result = function_call(**arguments)
   
       # Optionally transform the result
   
       return result
   
   
   @tool(tool_hooks=[confirmation_hook])
   def get_top_hackernews_stories(num_stories: int) -> Iterator[str]:
       """Fetch top stories from Hacker News.
   
       Args:
           num_stories (int): Number of stories to retrieve
   
       Returns:
           str: JSON string containing story details
       """
       # Fetch top story IDs
       response = httpx.get("https://hacker-news.firebaseio.com/v0/topstories.json")
       story_ids = response.json()
   
       # Yield story details
       final_stories = []
       for story_id in story_ids[:num_stories]:
           story_response = httpx.get(
               f"https://hacker-news.firebaseio.com/v0/item/{story_id}.json"
           )
           story = story_response.json()
           if "text" in story:
               story.pop("text", None)
           final_stories.append(story)
   
       return json.dumps(final_stories)
   
   
   # Initialize the agent with a tech-savvy personality and clear instructions
   agent = Agent(
       model=OpenAIChat(id="gpt-5-mini"),
       tools=[get_top_hackernews_stories],
       markdown=True,
   )
   
   agent.print_response(
       "Fetch the top 2 hackernews stories?", stream=True, console=console
   )
   
2. Run the example

   Install libraries

   pip install openai agno
   

   Export your key

   export OPENAI_API_KEY=xxx
   

   Run the example

   python hitl.py