External Tool Execution Stream

This example demonstrates how to execute tools outside of the agent using external tool execution with streaming responses. It shows how to handle external tool execution while maintaining real-time streaming capabilities.

Code

external_tool_execution_stream.py

"""🤝 Human-in-the-Loop: Execute a tool call outside of the agent

This example shows how to implement human-in-the-loop functionality in your Agno tools.
It shows how to:
- Use external tool execution to execute a tool call outside of the agent

Run `pip install openai agno` to install dependencies.
"""

import subprocess

from agno.agent import Agent
from agno.db.sqlite import SqliteDb
from agno.models.openai import OpenAIChat
from agno.tools import tool
from agno.utils import pprint


# We have to create a tool with the correct name, arguments and docstring for the agent to know what to call.
@tool(external_execution=True)
def execute_shell_command(command: str) -> str:
    """Execute a shell command.

    Args:
        command (str): The shell command to execute

    Returns:
        str: The output of the shell command
    """
    if command.startswith("ls"):
        return subprocess.check_output(command, shell=True).decode("utf-8")
    else:
        raise Exception(f"Unsupported command: {command}")


agent = Agent(
    model=OpenAIChat(id="gpt-5-mini"),
    tools=[execute_shell_command],
    markdown=True,
    db=SqliteDb(session_table="test_session", db_file="tmp/example.db"),
)

for run_event in agent.run(
    "What files do I have in my current directory?", stream=True
):
    if run_event.is_paused:
        for tool in run_event.tools_awaiting_external_execution:  # type: ignore
            if tool.tool_name == execute_shell_command.name:
                print(
                    f"Executing {tool.tool_name} with args {tool.tool_args} externally"
                )
                # We execute the tool ourselves. You can also execute something completely external here.
                result = execute_shell_command.entrypoint(**tool.tool_args)  # type: ignore
                # We have to set the result on the tool execution object so that the agent can continue
                tool.result = result

        run_response = agent.continue_run(
            run_id=run_event.run_id, updated_tools=run_event.tools, stream=True
        )  # type: ignore
        pprint.pprint_run_response(run_response)


# Or for simple debug flow
# agent.print_response("What files do I have in my current directory?", stream=True)

Usage

1. Create a virtual environment

   Open the Terminal and create a python virtual environment.

   Mac

   python3 -m venv .venv
   source .venv/bin/activate
   
2. Install libraries

   pip install -U agno openai
   
3. Export your OpenAI API key

   Mac/Linux

     export OPENAI_API_KEY="your_openai_api_key_here"
   
4. Create a Python file

   Create a Python file and add the above code.

   touch external_tool_execution_stream.py
   
5. Run Agent

   Mac

   python external_tool_execution_stream.py
   
6. Find All Cookbooks

   Explore all the available cookbooks in the Agno repository. Click the link below to view the code on GitHub:

   Agno Cookbooks on GitHub