All Products
Search

This Product

    Alibaba Cloud Linux:One-command deployment of Openclaw/Claude Code

    Document Center

    Alibaba Cloud Linux:One-command deployment of Openclaw/Claude Code

    Last Updated:Mar 28, 2026

    Background

    • OpenClaw is an open-source AI agent framework that connects to communication channels like DingTalk and large language models like Model Studio's Qwen through plugins.

    • Claude Code is an AI-powered coding assistant from Anthropic that lets you interact with your code using natural language in the terminal.

    • Installing and configuring these tools on an Agentic OS ECS instance can be challenging for several reasons:

      1. OpenClaw has an undocumented configuration file structure, DingTalk is not an officially supported channel, and the gateway requires a strict initialization sequence.

      2. The official Claude Code tutorial is for macOS and Ubuntu, which can cause compatibility issues and npm permission errors on Alinux.

      3. Documentation on how to configure the DashScope API key is missing for both tools.

    • With a Skill, the agent can complete the end-to-end deployment in a single command. This reduces a process that would otherwise be difficult to automate or require trial and error to just a few minutes.

    Use case 1: One-command deployment of OpenClaw

    Prerequisites

    • You have an Agentic OS ECS instance with at least 4 GB of RAM and internet access.

    • You have created an internal enterprise application on the DingTalk Developer Console and obtained the AppKey and AppSecret.

    Important: On the DingTalk Developer Console, enable the Bot Capability for your application and set the Message Reception Mode to Stream mode. You must also enable the following permissions: Card.Streaming.Write, Card.Instance.Write, and qyapi_robot_sendmsg.

    • You have obtained a DashScope API key from the Model Studio console.

    • You can enable YOLO mode to avoid manually confirming shell commands (/approval-mode yolo).

    Procedure

    Step 1: Run the deployment command

    1. In the Agentic OS terminal, send the following command to the agent:

    Install OpenClaw with AppKey <appkey>, AppSecret <appsecret>, and API key <dashscope apikey>

    1. The agent automatically performs the following actions:

      • Detects and installs the Node.js runtime environment.

      • Installs OpenClaw and the DingTalk plugin via npm, automatically adding the npmmirror registry to accelerate downloads in the Chinese mainland.

      • Initializes the OpenClaw project structure and configuration file.

      • Configures the gateway mode (gateway.mode).

      • Runs openclaw doctor --fix to generate an authentication token.

    Important: The gateway initialization order is critical. The gateway.mode configuration must be written before you run openclaw doctor --fix. The Skill has this sequence built-in, so no manual intervention is required.

    Step 2: Validate the deployment

    1. After the deployment is complete, find the bot in DingTalk and send it a test message.

    2. A correct reply from the bot confirms a successful end-to-end deployment.

    Notes

    • If the gateway fails to start during deployment, check that the initialization sequence is correct.

    • If the DingTalk bot is unresponsive, verify that all required application permissions, such as Card.Instance.Write and Card.Streaming.Write, are enabled.

    Use case 2: One-command deployment of Claude Code

    Prerequisites

    • You have an Agentic OS ECS instance with at least 4 GB of RAM and internet access.

    Important: Claude Code has a high memory footprint at runtime. Ensure your ECS instance has at least 4 GB of RAM.

    • You have obtained a DashScope API key from the Model Studio console (if you need to configure a third-party API).

    • You can enable YOLO mode to avoid manually confirming shell commands (/approval-mode yolo).

    Procedure

    Step 1: Run the installation command

    1. In the Agentic OS terminal, send the following command to the agent:

    Install Claude Code and configure the DashScope API key: <dashscope apikey>

    1. The agent automatically detects your system environment and selects the best installation method in the following order:

      • Native installation: Attempts to install using the system's native package manager first.

      • npm installation: If native installation is not available, it installs globally via npm and automatically fixes any EACCES permission issues.

      • nvm installation: If npm installation fails, it uses nvm to install an isolated Node.js environment before installing.

    Note: The installation process automatically falls back through these methods without manual intervention. The Skill handles compatibility issues on Alinux 4, including EACCES permission fixes.

    Step 2: Validate the installation

    The agent runs the following command to verify the installation:

    claude --version

    image.png

    Step 3: Configure the DashScope API key

    1. The agent automatically configures the DashScope API key for Claude Code.

    2. Start Claude Code and test that you can interact with the model.

    image.png

    Notes

    • Obtain your DashScope API key from the Alibaba Cloud Model Studio console and activate the necessary services in advance.

    • If all installation methods fail, check your network connectivity and Node.js version. We recommend using version 18 or later.