Agent

Load MCPToolbox hotel and booking toolsets for an interactive CLI agent, with a documented limitation on the manual-loading alternative.

Simple test script that connects to the MCP toolbox server

Use run_agent, the automatic toolsets= approach shown first. The manual-loading alternative in this source excerpt fails because an unfiltered MCPToolbox does not initialize its core client. Its sample auth sources and bound parameters are also absent from the demo configuration. Do not enable that alternative.

agent.py
"""
Simple test script that connects to the MCP toolbox server
"""

import asyncio
from textwrap import dedent

from agno.agent import Agent
from agno.models.openai import OpenAIChat
from agno.tools.mcp_toolbox import MCPToolbox

# ---------------------------------------------------------------------------
# Create Agent
# ---------------------------------------------------------------------------


url = "http://127.0.0.1:5001"


async def run_agent(message: str) -> None:
    """Run an interactive CLI for the Hotel agent with the given message."""

    # Approach 1: Load specific toolset at initialization
    async with MCPToolbox(
        url=url, toolsets=["hotel-management", "booking-system"]
    ) as db_tools:
        # returns a list of tools from a toolset
        agent = Agent(
            model=OpenAIChat(),
            tools=[db_tools],
            instructions=dedent(
                """ \
                You're a helpful hotel assistant. You handle hotel searching, booking and
                cancellations. When the user searches for a hotel, mention it's name, id,
                location and price tier. Always mention hotel ids while performing any
                searches. This is very important for any operations. For any bookings or
                cancellations, please provide the appropriate confirmation. Be sure to
                update checkin or checkout dates if mentioned by the user.
                Don't ask for confirmations from the user.
            """
            ),
            markdown=True,
        )

        # Run an interactive command-line interface to interact with the agent.
        await agent.acli_app(input=message, stream=True)


async def run_agent_manual_loading(message: str) -> None:
    """Alternative approach: Manual loading with custom auth parameters."""

    # Approach 2: Manual loading with custom auth parameters
    async with MCPToolbox(url=url) as toolbox:  # No filter parameters
        # Load specific toolsets with custom auth
        hotel_tools = await toolbox.load_toolset(
            "hotel-management",
            auth_token_getters={"hotel_api": lambda: "your-hotel-api-key"},
            bound_params={"region": "us-east-1"},
        )

        booking_tools = await toolbox.load_toolset(
            "booking-system",
            auth_token_getters={"booking_api": lambda: "your-booking-api-key"},
            bound_params={"environment": "production"},
        )

        # Combine tools as needed
        selected_tools = []
        selected_tools.extend(hotel_tools)
        selected_tools.extend(booking_tools[:2])  # Only first 2 booking tools

        agent = Agent(
            tools=selected_tools,
            instructions=dedent(
                """ \
                You're a helpful hotel assistant. You handle hotel searching, booking and
                cancellations. When the user searches for a hotel, mention it's name, id,
                location and price tier. Always mention hotel ids while performing any
                searches. This is very important for any operations. For any bookings or
                cancellations, please provide the appropriate confirmation. Be sure to
                update checkin or checkout dates if mentioned by the user.
                Don't ask for confirmations from the user.
            """
            ),
            markdown=True,
            add_history_to_context=True,
        )

        await agent.acli_app(input=message, stream=True)


# ---------------------------------------------------------------------------
# Run Agent
# ---------------------------------------------------------------------------

if __name__ == "__main__":
    # Use the original approach
    asyncio.run(run_agent(message=""))

    # Or use the manual loading approach
    # asyncio.run(run_agent_manual_loading(message=None))

Run the Example

Set up your virtual environment

uv venv --python 3.12
source .venv/bin/activate

Install dependencies

uv pip install -U "agno[mcp]" openai toolbox-core

Export your OpenAI API key

export OPENAI_API_KEY="your_openai_api_key_here"

Clone Agno

Clone the pinned Agno source and run the remaining commands from its root:

git clone https://github.com/agno-agi/agno.git
cd agno
git checkout 8f36eaf2d18e91afa7b327eec66a3cd3685dcb87

Start MCP Toolbox

Install and start Docker with Compose support. This local demonstration starts PostgreSQL on host port 5432 and Toolbox on 5001; both ports must be available. Start the services:

cd cookbook/91_tools/mcp/mcp_toolbox_demo
docker compose up -d
cd ../../../..

Run the example

Run the example from the repository root:

python cookbook/91_tools/mcp/mcp_toolbox_demo/agent.py

Full source: cookbook/91_tools/mcp/mcp_toolbox_demo/agent.py