Skip to main content
The Function class represents a tool that can be called by an agent. It includes the function definition, parameters, hooks, and execution logic.

​
Core Fields

​
name
str
required
Name of the function (must be a-z, A-Z, 0-9, underscores, or dashes, max 64 chars).
​
description
str
default:"None"
Description of what the function does, used by the model to choose when to call it.
​
parameters
Dict[str, Any]
JSON Schema object describing function parameters.
​
strict
bool
default:"None"
Enable strict schema validation (for structured outputs).
​
entrypoint
Callable
default:"None"
The actual function to be called.

​
Instructions

​
instructions
str
default:"None"
Additional instructions for using the function.
​
add_instructions
bool
default:"True"
If True, adds instructions to the agent’s system message.

​
Display & Control

​
show_result
bool
default:"False"
If True, the function call result is shown along with sending it to the model.
​
stop_after_tool_call
bool
default:"False"
If True, the agent stops after executing this function.

​
Hooks

​
pre_hook
Callable
default:"None"
Hook that runs before the function is executed.
​
post_hook
Callable
default:"None"
Hook that runs after the function is executed (success or failure).
​
tool_hooks
List[Callable]
default:"None"
List of hooks to run around tool calls as middleware.

​
HITL (Human-in-the-Loop)

​
requires_confirmation
bool
default:"None"
If True, requires user confirmation before execution.
​
requires_user_input
bool
default:"None"
If True, requires user input before execution.
​
user_input_fields
List[str]
default:"None"
List of fields that should be provided by the user.
​
external_execution
bool
default:"None"
If True, the function will be executed outside the agent’s control.
​
approval_type
str
default:"None"
Approval type: “required” (blocking) or “audit” (non-blocking).

​
Caching

​
cache_results
bool
default:"False"
Enable result caching.
​
cache_dir
str
default:"None"
Directory to store cache files.
​
cache_ttl
int
default:"3600"
Cache time-to-live in seconds.

​
Methods

​
from_callable()

Create a Function from a callable. 
from agno.tools.function import Function

def my_tool(query: str) -> str:
    """Search for information."""
    return f"Results for {query}"

func = Function.from_callable(my_tool)

​
to_dict()

Serialize function to dictionary. 
data = function.to_dict()

​
from_dict()

Deserialize function from dictionary. 
function = Function.from_dict(data)

​
FunctionCall

Represents an instance of a function being called.

​
Methods

​
execute()

Execute the function call synchronously. 
result = function_call.execute()

​
aexecute()

Execute the function call asynchronously. 
result = await function_call.aexecute()

​
Example Usage

from agno.tools.function import Function

def get_weather(city: str) -> str:
    """Get the weather for a city.
    
    Args:
        city: The name of the city
    """
    return f"Weather in {city}: Sunny, 72°F"

func = Function.from_callable(get_weather)

Build docs developers (and LLMs) love

Get started for freeTalk to us