AgentBay Dify Plugin

中文文档 | 日本語

Introduction

AgentBay is a cloud sandbox service from Alibaba Cloud WuYing, providing isolated computing environments (Linux, Windows, Browser, etc.) that enable AI Agents to safely execute code, manipulate files, automate browsers, and perform complex tasks.
For detailed information and usage instructions about AgentBay, please refer to the official website: https://www.aliyun.com/product/agentbay

Why This Plugin:

Traditional AI Agents are limited to text-based interactions and cannot execute actual computational tasks. This plugin integrates AgentBay to give your Dify applications "real action capabilities":

🔧 Code Execution: Run Python, Shell scripts in isolated environments for data analysis, file conversion, etc.
🌐 Browser Automation: Automate web browsing, form filling, data scraping, and web operations
📁 File Operations: Read/write cloud files, process documents and logs
🖥️ Desktop Automation: Control Windows applications for RPA workflows
☁️ Cloud-Based: All operations run in the cloud with security isolation, no local environment needed

Typical Use Cases:

Data Analysis Agent: Run Python scripts to analyze user-uploaded data
Web Scraping Agent: Automatically visit websites to extract information
Automated Testing Agent: Execute web application automated tests
Document Processing Agent: Batch convert and process file formats
RPA Office Agent: Automate repetitive desktop operations

Features

Session Management

Create various cloud environments (Linux, Browser, Code, Windows, Mobile)
View and manage all active sessions
Safely delete sessions and clean up resources

Command & Code Execution

Execute Shell commands in cloud environments
Run Python and other programming languages
Configurable execution timeout

File Operations

Read and write files
List directory contents

Browser Automation

Web navigation, element interaction (click, type, scroll)
Page screenshots, content extraction
Element analysis, wait for loading

Desktop UI Automation

Desktop screenshots
Mouse clicks, keyboard input

Quick Start

Get API Key

Visit AgentBay Console to get your API key.

Configure Plugin

After installing the plugin in Dify, configure the API Key parameter.

Basic Usage

1. Create Session

2. Perform Operations

Execute commands:

File operations:

Browser automation:

3. Clean Up Resources

Demo Examples

We provide ready-to-use demo examples that showcase the plugin's capabilities. These examples are available in the demo directory.

You can quickly get started by importing these examples into Dify:

Download the demo file (in Dify DSL format) from the demo directory
In Dify, go to "Studio" and click "Import DSL File"
Select the downloaded demo file to import
Configure the AgentBay API Key in the imported workflow
Run the demo to see how the plugin works

Tools Reference

session_create

Create a new cloud environment session. Supports 5 environment types: (default), (browser), (development), (Windows), (mobile). Returns session_id for subsequent operations.

session_list

List all active sessions created by this plugin under the current account, including session ID, environment type, etc.

session_delete

Delete specified session and release cloud resources. Recommended to clean up promptly after operations.

command_execute

Execute Shell commands in a session with support for custom working directory and timeout. Suitable for command-line operations in Linux/Windows environments.

code_execute

Run code in a session, supporting Python and other programming languages. Can specify working directory, suitable for data processing, automation scripts, etc.

file_operations

Unified file operation tool, specify operation type via parameter:

read: Read file content
write: Write to file (automatically creates non-existent files)
list: List directory contents

browser_automation

Full-featured browser automation tool, requires environment. Supports multiple operations via parameter:

navigate: Visit web pages
click/type: Interact with page elements (locate using CSS selectors)
scroll: Page scrolling
screenshot: Capture page screenshots (supports full page or viewport)
get_content: Get page HTML content
analyze_elements: Analyze page structure to help find correct selectors
wait_element/wait: Wait for elements or delays

ui_operations

Desktop UI automation tool, suitable for Windows, browser and other graphical environments. Supports via parameter:

screenshot: Capture desktop screenshots
click: Click at specified coordinates
type: Type text
key: Press specified keys (e.g., Enter, Tab, etc.)

Use Cases

Web Automation Testing

Create browser_latest session
Use browser_automation to navigate and interact
Screenshot to verify results
Delete session

Data Processing

Create code_latest session
Use file_operations to upload data files
Use code_execute to run analysis scripts
Use file_operations to download results
Delete session

System Operations

Create linux_latest session
Use command_execute to run check commands
Use file_operations to read logs
Delete session

Contact

GitHub Issues
Email: [email protected]

Disclaimer

This plugin is provided as a technical tool. Please note the following when using:

Legal Use: Ensure all operations comply with laws and regulations, and do not use for illegal purposes
Data Responsibility: Users are responsible for the data they process, and it is recommended to avoid processing sensitive information
Service Dependency: Must comply with Alibaba Cloud WuYing AgentBay service terms
Risk Assumption: Usage involves risks such as data loss and service interruption, which users assume at their own risk
Liability Limitation: Developers are not responsible for any losses caused by usage

Using this plugin indicates agreement to the above terms.