mcpField
A
Agentbay

Agentbay

Remote
github.com/Michael98671Cloud & Infra7Updated 47w ago
Repository

About Agentbay

Wuying AgentBay MCP Server connects AI agents to Alibaba Cloud’s serverless execution environments. It offers isolated, persistent user state and browser, file, and terminal tools for enterprise and developer tasks.

Key features

  1. 01Serverless cloud execution environments
  2. 02Isolated, persistent user configurations and files
  3. 03Browser, file, and terminal tools
  4. 04Real-time edge-cloud interaction
  5. 05SDK, API, and MCP integration

Use cases

  1. 01Run browser tasks in a cloud environment
  2. 02Execute file and terminal operations for an agent
  3. 03Retain user configuration across sessions

MCP support

It exposes browser, file, and terminal tools through an MCP server. Connections support SSE and STDIO.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://agentbay.wuying.aliyuncs.com/sse?APIKEY=akm-ea43de3f-72e0-4ef8-bf67-fa3222b373bc&IMAGEID=windows_latest
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Tools

Details

From the project README

Wuying AgentBay MCP Server

English | 中文

License: MIT

Overview

Alibaba Cloud's Wuying AgentBay is a cloud infrastructure designed for Agents in the AI era. It provides enterprises, developers, and AI vendors with a one-click configuration tool and execution environment for AI Agent tasks. Through the Wuying API or AgentBay MCP Server, Agent applications can quickly integrate and utilize relevant tools, running serverlessly in the cloud environment.

What is Wuying?

Wuying (无影), also known as Alibaba Cloud Workspace, is Alibaba Cloud's innovative cloud-network fusion computing architecture that provides secure and efficient computing experiences through cloud computers, cloud applications and cloud terminals. It features:

  • Strong security
  • Zero maintenance
  • Light asset management
  • Easy integration
  • Global deployment with low-latency access

What is AgentBay?

AgentBay is Wuying's cloud infrastructure specifically designed for AI Agents. It provides:

  • One-click configurable AI Agent task execution tools
  • Secure execution environments
  • Enterprise-grade infrastructure for developers and AI vendors
  • Quick integration through SDK and MCP Server
  • Serverless capabilities
Key Features
  1. Standard Runtime Environment

    • Pre-integrated standard tools for Agent task execution
    • MCP (Model Context Protocol) encapsulation
    • Rapid enterprise integration capabilities
  2. User State Persistence

    • Secure isolation of user configurations and cookies
    • Dynamic mounting of user states
    • Cloud environment closely mimicking local setup
  3. Real-time Edge-Cloud Interaction

    • Proprietary ASP protocol for real-time edge-cloud streaming
    • Local device awareness and control
    • Network and peripheral redirection channels

Technical Architecture

  • Persistent System: Custom persistent memory to retain user states and files
  • Edge-Cloud Communication: Proprietary ASP protocol for real-time communication between the cloud and terminal devices
  • Infrastructure: Provides the ability to scale to 10,000 concurrent operations in seconds, based on Alibaba Cloud's resource pool
  • Service Model: Serverless capabilities with one-click environment session management
  • Tool Integration: Browser, File, Terminal, and other standard MCP tools
  • Integration Options: SDK+API or MCP Server compatibility

Cloud Service Specifications

Cloud Resources

The service supports calling shared resources, with limited free service during the public beta, and a concurrency limit of 10 instances. Service regions are automatically assigned based on customer access IP.

Runtime Environment

The service supports built-in Linux version images. Wuying continuously enriches and updates image capabilities while ensuring backward compatibility.

MCP Tools

The initial release environment supports the Browser tool, File tool, and Terminal tool by default. For more updates on tools, please visit the official site.

Getting Started

Step 1: Create an API Key

Please create your own API Key.

  1. Log in to the AgentBay Console.
  2. Click Service Management in the left navigation bar.
  3. On the Service Management page, click Create API KEY.
  4. Enter a name in the Create API KEY dialog box and click OK.

Note: During the public beta, only 10 API Keys can be created. The created API Key will be displayed in the list on the service management page, and you can copy it for later use.

Step 2: Complete MCP Service Configuration

Confirm the runtime environment and obtain MCP information.

  1. Log in to the AgentBay Console.
  2. Click Configure Resources in the left navigation bar.
  3. Select an API Key you created from the API Key dropdown list.
  4. Select an image from the Select Image dropdown list.

Note: The service supports built-in Linux version images. Wuying continuously enriches and updates image capabilities while ensuring backward compatibility.

  1. On the right side of Tool Capabilities and Interface, check available MCP tools. Click View to see more details.
  2. In the Obtain MCP Information text box, click Copy Code in the top right corner, and add the code to tools that support MCP (such as Cline, Cursor, etc.).

Here's an example using Cursor V0.48.9:

  • Open the Cursor Settings panel and click MCP in the left navigation bar.
  • In the MCP Servers panel, click Add new global MCP server in the top right corner.
  • Paste the following code into the opened mcp.json file and save it. Supported installation methods include SSE and STDIO.

Note: Please replace YOUR_API_KEY with the API Key starting with akm- created in Step 1.

Here's an example using the SSE method: The SSE method completes a handshake through an online long connection each time the service is run, without requiring plugin downloads.

{
  "mcpServers": {
    "wuying_mcp_server": {
      "url": "https://agentbay.wuying.aliyun.com/sse?APIKEY=YOUR_API_KEY"
    }
  }
}

After the task execution is completed, the cloud environment may be released. If you wish to retain related configuration or environment for next time, add the EXTERNALID parameter in the code above. Example:

{
  "mcpServers": {
    "wuying_mcp_server": {
      "url": "https://agentbay.wuying.aliyun.com/sse?APIKEY=YOUR_API_KEY&EXTERNALID=user001"
    }
  }
}
Step 3: Using Screen Streaming (Optional)

Support is provided for opening cloud computer screens through the Wuying Web SDK embedded in an iframe or by pulling the Wuying Web client via a browser. When calling the cloud computer screen stream, MCP will return a link with login parameters, formatted as https://wuying.aliyun.com?mcp.html?authcode=<authCode>&resourceId=<resourceId>.

  • This link has validity and open limits, and it is recommended to use it immediately.
  • You can configure keyboard and mouse interaction switches through the parameters &input=true&keyboard=true in the URL, or configure interaction switches in real time through the Web SDK interface. For detailed information, please refer to Web SDK Documentation。

Next Steps

You can test the effect using built-in prompts like open the browser with wuying-agentbay and access wuying.aliyun.com or similar prompts. If configured correctly, the Wuying cloud environment will be initiated, and browser operations will be executed. During this process, the model will provide a Web link for you to open the Wuying screen stream and take over operations. Next Steps

License

MIT

Support

For support and more information, please visit:

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.