For a long time, traditional AI applications have been stuck in an awkward predicament: despite their powerful capabilities in understanding and reasoning, they remain isolated from the real-world operational control loop. It’s like a brilliant consultant who can only “talk the talk but not walk the walk”: capable of analyzing data, drafting reports, and offering recommendations, yet unable to press a single execution button himself. When users need to turn AI’s analysis into concrete action, that “last mile” gap still requires human effort to bridge—you still have to manually open another system, enter commands, and confirm execution. The emergence of MCP server functions is fundamentally changing this landscape. As a “universal connector” in the AI field, MCP provides AI models with a standardized way to discover, access, and understand real-time data scattered across different systems. And when this unified context-aware capability is deeply integrated with Awesun remote control technology, a qualitative leap occurs: Awesun AweSun MCP enables AI to, for the first time, simultaneously possess “eyes that perceive the world” and “hands that intervene in the world.”

1. Awesun what AweSun MCP can do

Professional remote control capability: Directly encapsulates the Awesun mature remote control system, so that AI remote control also has stability, safety and compatibility.

“Visualized” execution of automated operations and maintenance: AI enables the automated execution of tasks such as fault diagnosis and performance optimization. By integrating the “visual recognition and automated operations” capabilities of external large models, AI can “operate whatever it sees”—capturing screenshots to identify interface elements, parsing the current state, and then automatically triggering the corresponding actions, completing intelligent desktop automation in three steps and significantly boosting operational efficiency. Moreover, compared with command-line operations, the “visual desktop” approach makes every step of AI’s operation traceable. For example: simply send the text command “screenshot” to the AI client, and it will automatically capture the screen of the current remote device and save it.

"Conversational" multi-device management: Simply tell the AI your needs, and it will automatically handle tasks such as device discovery. AI can quickly locate the target device by name or status, eliminating the need to manually scroll through lists; all operations are easily handled via natural language, truly enabling “control multiple devices with just a few words.” For example: Simply send the text command “Remotely restart the server at 192.168.1.100” to the AI client, and the AI client will automatically perform a shutdown and restart, with no need for on-site human intervention throughout the process.

Seamless Cross-Platform Collaboration: Fully supports both Windows and macOS, meeting remote control needs across different platforms. Developers can write code on a local Windows machine and remotely manage applications deployed on Linux servers via the MCP protocol, while operations personnel can use macOS to simultaneously oversee both Windows virtual machines and cloud resources. The MCP protocol serves as a unified bridge, ensuring “initiate once, respond everywhere,” enabling cross-platform collaboration to be as seamless as silk.

2. Pre-conditions

① In the host environment, upgrade the Awesun client to version V16.2.3 or later (click to download) ② Install an AI client that supports MCP (such as OpenCode, Cherry Studio, etc.)

3. Operation steps

Deploying AweSun MCP Server 3.1

Operation path: [Navigation] -> [Awesun MCP] (1) In the host control environment, open the Awesun software and enable “Enable MCP Server”; (2) Select the appropriate service type based on the usage scenario: ①Stdio (recommended): Local process communication, suitable for AI clients running locally ②Streamable HTTP: Remote communication based on HTTP, suitable for scenarios that require cross-network calls. (3) After making your selections, copy the generated configuration information for the AI client you need to connect to at present and set it aside.

3.2 Configure AI Client

Using OpenCode, Claude Code, and Cherry Studio as examples, we will demonstrate the configuration process. We recommend using OpenCode for a better user experience.

3.2.1 OpenCode

Demonstrating the configuration process using OpenCode as an example. For users with no programming experience, we recommend downloading the OpenCode desktop version (download link).

3.2.1.1 Configuration Model Service

Operation path: [Settings] -> [Server] -> [Provider] (1) Run the OpenCode desktop application and select a model provider to connect to. (2) After setting up the supplier, activate the corresponding model; we recommend using large models with visual understanding capabilities, such as Kimi K2.5 or Gemini 2.5 Pro.

(3) To better understand AI’s current operational status, it is recommended to enable the following settings.

3.2.1.2 Configuring the MCP Server

3.2.1.2.1 New Work Area

Create a new folder as the Opencode workspace, then initialize the project according to the structure shown below.

awesun-mcp-example           // 工作区目录
    ├── AGENTS.md            // 系统提示词
    └── opencode.json        // 工作区配置文件

3.2.1.2.2 Load Workspace

Open the workspace directory using OpenCode (make sure to select the awesun-mcp-example folder).

3.2.1.2.3 Configure Workspace Prompt Words

Manually edit the awesun-mcp-example/AGENTS.md memory file and configure the default system prompt. Reference link: AweSun MCP Example Prompts

3.2.1.2.4 Configuring the Workspace MCP Server

Manually edit the awesun-mcp-example/opencode.json configuration file, and paste the MCP configuration information copied from the Awesun client into the file. A complete example is as follows:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "serverName": {
      "type": "local",
      "command": [
    "/Applications/AweSun.app/Contents/Helpers/awesun-mcp-server"
      ],
      "enabled": true,
      "environment": {
        "AWESUN_API_URL": "http://127.0.0.1:8980",
        "AWESUN_API_TOKEN": "xxxxxxx"
      }
    }
  }
}

3.2.1.3 Checking Server Connection Status

After writing the configuration file, you must restart the OpenCode client for the changes to take effect. Then, follow these steps to verify the connection status: (1) Restart the OpenCode client; (2) Enter the command /mcp in the input box to check the MCP server connection status.

As shown above, if “awesun-mcp-server” is displayed as connected and its service status is set to “running,” the configuration is complete.

3.2.1.4 Function Demonstration

Once everything is ready, verify the functionality by following these steps: (1) Enter /new in the input box to start a new chat session. (2) Enter a message: Help me query the configuration information of xxx devices in the device list

3.2.2 Claude Code

3.2.2.1 Installing the CLI Tools

Taking Windows 11 as an example, open the terminal, create a new Windows PowerShell window, and run the following command: irm https://claude.ai/install.ps1 | iex

3.2.2.2 Configuring the MCP Server

3.2.2.2.1 New Work Area

Create a new folder on the computer as the workspace for the claude code, and initialize the project as shown below. awesun-mcp-example-claude // workspace directory ny_CLAUDE. md // memory file . mcp.json // MCP configuration file

3.2.2.2.2 Configure Workspace Prompt Words

Manually edit the awesun-mcp-example-claude/CLAUDE.md memory file and set the system prompt. Reference link: AweSun MCP Example Prompts

3.2.2.2.3 Configuring the Workspace MCP Server

Manually edit the awesun-mcp-example-claude/.mcp.json configuration file, and paste the MCP configuration information copied from the Awesun client into the file. A complete example is as follows:

{  
    "mcpServers": {  
        "awesun-mcp-server": {  
            "command": "C:\Program Files\Oray\AweSun\flutter\awesun-mcp-server.exe",  
            "env": {  
                "AWESUN_API_URL": "http://127.0.0.1:8908",  
                "AWESUN_API_TOKEN": "MTMxNjU1xxxxxxxxxxxxxxxxxxYTktMDkzMmVjN2IyY2Y2"  
            }  
        }  
    }  
}  

3.2.2.3 Configure model services

Preparations are now largely complete; the next step is to integrate the large language model. Please follow these steps: Open File Explorer, right-click on “This PC,” select “Properties,” go to “Advanced system settings,” and then click “Environment Variables.” Create the relevant environment variables according to the following table.

3.2.2.3.1 Load Workspace

(1) In File Explorer, locate the workspace directory you just created and open it. (2) In the File Explorer address bar, type the command claude and press Enter to load the Claude terminal. Note: The first time you enter, you’ll need to confirm; simply select “Yes.”

3.2.2.3.2 Check Work Area Status

In the Claude terminal, enter the command /status and press Enter to view the output. If you see “Anthropic base URL: https://api.kimi.com/coding,” it indicates that the environment variable has been configured successfully. If awesun-mcp-server appears ·✔This indicates that the MCP server has been configured successfully.

3.2.2.4 Function Demonstration

Please follow these steps to verify: (1) If you are currently on the status page, press the ESC key in the Claude terminal to return to the home page until the output box appears. (2) Input message: Please help me query the configuration information of device xxx in the device list.

Note: By default, Claude‑Code requires manual user confirmation before invoking each tool. To enable fully automated execution, you can re-enter the terminal using the command claude --dangerously-skip-permissions.

3.2.3 Cherry Studio

Generally not recommended, because the implementation effect is not as good as the first two. However, if you want a simple experience and think the above configuration is too complicated, you can try Cherry Studio.

3.2.3.1 Import MCP Configuration

Operation path: [Settings] -> [MCP Server] (1) Open Cherry Studio, then go to 【Settings】→【MCP Server】; (2) Click “Add” and select “Import from JSON”; (3) Paste the copied Awesun server configuration information into the input box and click OK. (4) After the import is successful, ensure that the MCP server status is set to “Enabled.”

3.2.3.2 Configuration Model Service

Operation path: [Settings] -> [Model Service] Enable a model provider as the underlying model for testing MCP functionality (such as Alibaba Cloud’s Bailing model).

3.2.3.3 Test the MCP function

Cherry Studio comes with a default general-purpose assistant that can be used for basic MCP functionality testing.

3.2.3.3.1 Configure Generic Assistant

(1) Return to the Cherry Studio homepage, select the default assistant and edit; (2) Set the MCP server to manual (or automatic) mode, and select awesun-mcp-server. (3) Model setting Preset model: It is recommended to select a model that supports visual recognition capabilities so that the model can analyze the content of screenshots. If the visual function is not required, a normal text inference model is sufficient. Adjust model parameters: To prevent context loss, it is recommended to set the context length to 10 or higher.

3.2.3.3.2 Setting Prompt Words

To help the large model better understand the purpose of the MCP tool, it is recommended to configure preset prompts: Reference link: AweSun MCP Example Prompts

3.2.3.4 Function Demonstration

After completing the above configuration, you can enter commands in the chat interface to perform a test. For example: “Check the game console’s system information” or “Initiate remote control of the game console.” AI will invoke the corresponding MCP tools based on your instructions, enabling automated operations.

4.Q&A

(1) Why did my AI remote control fail?
For device security reasons, remote control requires verification of the access password or system account password; AI cannot automatically obtain your password. You only need to manually establish remote control of the device once; afterward, the AI will be able to handle remote control for you as usual.

(2) Why didn’t my AI succeed in its operations on the remote desktop? Desktop automation requires robust image‑recognition capabilities. We recommend switching to a larger model with stronger visual intelligence, such as Kimi‑2.5, to achieve a higher success rate.

5. Appendix: AweSun MCP Tool Documentation

Awesun MCP Server provides Awesun client operation capabilities, including three major categories of tools: device management, remote control sessions, and desktop operations. (View details)

6. Advanced play

(1) Some scenario-specific skills are used as examples; by streamlining workflows, AI can enhance the efficiency and accuracy of task execution, thereby covering certain functions traditionally handled by RPA. (View details)

(2) Based on the Awesun MCP service, it provides tools for progressive disclosure of API calls to Claude Code, Open Code, OpenClaw, and other AI Agents that support Skills. (View details)