daytona-adk

Google ADK plugin for secure code execution in Daytona sandboxed environments. This plugin enables ADK agents to execute Python, JavaScript, and TypeScript code, run shell commands, upload scripts and datasets, and read outputs from files in isolated sandboxes.

Installation

pip install daytona-adk

Configuration

You must configure credentials for both Daytona (sandbox infrastructure) and Google Gemini (the LLM model used by ADK agents).

Daytona API Key

Get your API key from Daytona Dashboard.

You can configure it in one of three ways:

1. Set the DAYTONA_API_KEY environment variable:

   export DAYTONA_API_KEY="your-daytona-api-key"
   

2. Add it to a .env file in your project root:

   DAYTONA_API_KEY=your-daytona-api-key
   

3. Pass the API key directly when instantiating DaytonaPlugin:

   from daytona_adk import DaytonaPlugin
   
   plugin = DaytonaPlugin(api_key="your-daytona-api-key")
   

Google API Key (for Gemini LLM)

Get your API key from Google AI Studio.

This key is required to access Google's Gemini models (e.g., gemini-2.0-flash, gemini-1.5-pro).

You need to set it as an environment variable:

export GOOGLE_API_KEY="your-google-api-key"

Or add it to your .env file:

GOOGLE_API_KEY=your-google-api-key

Quick Start

import asyncio
from google.adk.agents import Agent
from google.adk.runners import InMemoryRunner
from daytona_adk import DaytonaPlugin


async def main():
    # Create the Daytona plugin
    plugin = DaytonaPlugin()

    # Create an agent with Daytona tools
    agent = Agent(
        model="gemini-2.0-flash",
        name="agent_with_sandbox",
        tools=plugin.get_tools(),
    )

    # Run with InMemoryRunner
    async with InMemoryRunner(
        app_name="my_app",
        agent=agent,
        plugins=[plugin],
    ) as runner:
        response = await runner.run_debug(
            "Execute this Python code: print('Hello from Daytona!')"
        )
        print(response)


if __name__ == "__main__":
    asyncio.run(main())

Usage Patterns

Using with App (Recommended)

from google.adk.agents import Agent
from google.adk.apps import App
from google.adk.runners import InMemoryRunner
from daytona_adk import DaytonaPlugin

plugin = DaytonaPlugin()

agent = Agent(
    model="gemini-2.0-flash",
    name="agent_with_sandbox",
    tools=plugin.get_tools(),
)

app = App(
    name="my_app",
    root_agent=agent,
    plugins=[plugin],
)

async with InMemoryRunner(app=app) as runner:
    response = await runner.run_debug("Execute Python: print('Hello!')")

Using with Runner Directly

from google.adk.agents import Agent
from google.adk.runners import InMemoryRunner
from daytona_adk import DaytonaPlugin

plugin = DaytonaPlugin()

agent = Agent(
    model="gemini-2.0-flash",
    name="agent_with_sandbox",
    tools=plugin.get_tools(),
)

async with InMemoryRunner(
    app_name="my_app",
    agent=agent,
    plugins=[plugin],
) as runner:
    response = await runner.run_debug("Run command: echo 'Hello!'")

Available Tools

The plugin provides the following tools to ADK agents:

Tool Parameters

execute_code_in_daytona

• code (required): The code snippet to execute
• language (required): Programming language - python (default), javascript, or typescript
• env: Environment variables as key-value pairs
• argv: Command line arguments
• timeout: Timeout in seconds

execute_command_in_daytona

• command (required): The shell command to execute
• cwd: Working directory
• env: Environment variables
• timeout: Timeout in seconds

upload_file_to_daytona

• file_path (required): Destination path for the file
• content (required): File content to upload

read_file_from_daytona

• file_path (required): Path of the file to read

start_long_running_command_daytona

• command (required): The command to start
• timeout: Timeout in seconds

Plugin Configuration

DaytonaPlugin(
    api_key="your-api-key",           # Daytona API key (or use DAYTONA_API_KEY env var)
    plugin_name="daytona_plugin",     # Plugin identifier
    sandbox_name="my-sandbox",        # Optional sandbox name
    env_vars={"KEY": "value"},        # Environment variables for the sandbox
    labels={"env": "dev"},            # Custom labels
    auto_stop_interval=15,            # Auto-stop after N minutes (0 to disable)
    auto_delete_interval=60,          # Auto-delete stopped sandbox after N minutes
)

Examples

See the examples/ directory for complete working examples:

Usage Patterns:

• patterns/with_app.py - Using DaytonaPlugin with the App pattern (recommended)
• patterns/with_runner.py - Using DaytonaPlugin with Runner directly

Specific Use Cases:

• multi_language_execution.py - Execute Python, JavaScript, and TypeScript code
• file_operations.py - Upload scripts, execute them, and read output files
• long_running_process.py - Start and manage background processes

Architecture

packages/adk-plugin/
├── daytona_adk/
│   ├── __init__.py           # Package exports
│   ├── plugin.py             # DaytonaPlugin (extends BasePlugin)
│   └── tools.py              # ADK tool implementations
├── examples/
│   ├── patterns/
│   │   ├── with_app.py       # App pattern example
│   │   └── with_runner.py    # Runner pattern example
│   ├── multi_language_execution.py  # Multi-language code execution
│   ├── file_operations.py           # Script upload and file reading
│   └── long_running_process.py      # Background process management
├── tests/
│   └── test_tools.py         # Tool unit tests
└── pyproject.toml

Key Components

1. DaytonaPlugin: Extends BasePlugin

   • Creates and manages sandbox lifecycle
   • Provides tools via get_tools()
   • Implements lifecycle callbacks

2. Tools: ADK tool implementations using BaseTool

   • Each tool wraps Daytona SDK operations
   • Shared sandbox instance for state persistence

License

Apache 2.0 - See LICENSE file for details.

References

• Daytona Documentation
• Google ADK Documentation