Docs

Elastic Compute Service

Quickly deploying OpenClaw via marketplace images

Quickly deploying OpenClaw on Windows and integrating Feishu AI assistant

OpenClaw is an open-source, self-hosted AI agent tool designed with a local-first philosophy. It focuses on proactive execution and cross-platform interaction, with the core goal of enabling AI to complete real-world tasks through natural language instructions. This topic describes how to quickly access and use OpenClaw with simple configuration through a marketplace image predeployed with OpenClaw. You can later integrate OpenClaw with your existing Feishu application to build your own personal AI assistant.

tip

This feature is currently only supported for instances using the Subscription billing method. If your existing instance uses the Pay-as-you-go billing method, you may either change the billing method or deploy OpenClaw manually.

Image information

Image architecture: x86_64

Operating system: Windows 2022 (Chinese edition)

Latest version: 2026.2.9

Limits

Instance type family

The current image supports the following instance families. Choose based on your needs.

Type	Instance type family
General-purpose	g4i
		g4il
		g4al
		g3il
		g3al
Compute-optimized	c4i
		c4il
		c4al
		c3il
		c3al
Shared	e

Instance configuration

To balance the running requirements of OpenClaw and the overhead of the Windows operating system, and to prevent service interruptions and poor user experience caused by insufficient resources, we strongly recommend that you select an instance with 2 vCPU, 4 GiB memory or higher.

tip

If your instance configuration is lower than 2 vCPU and 4 GiB memory, you can change the instance type for a higher configuration.

Procedure

Step 1: Deploy OpenClaw

Newly purchased ECS instances

Existing ECS instances

If you do not have an ECS instance or want to deploy OpenClaw on a new instance, you can refer to this section to purchase a new instance from a marketplace image and deploy OpenClaw.

In the left-side navigation pane, select "Instances & Images > Instances".

In the top menu bar, select the region and project.

tip

After you select a project, ECS instances that you create can be associated only with the selected project. If you select "All Resources", ECS instances that you create can be associated with any project under the current account.

Select "Create instance > Custom purchase" to start purchasing an ECS instance. For more information about instance configuration, refer to Purchasing an instance by following the wizard.

warning

Select "Subscription" as the billing method.

When you configure the parameter "Image", select "Marketplace Image" and choose "OpenClaw2.9-Windows".

Step 2: Activate ModelArk and obtain the configuration

If you have activated the ModelArk service and obtained an API Key, skip this step. Otherwise, refer to this section for guidance on activating the service and obtaining an API Key.

In the directory tree on the left, select "Configure > API keys" to create and obtain an API Key.

tip

The obtained API Key will be used to configure OpenClaw. Keep it secure.

In the directory tree on the left, select "GenAI studio > Foundation models", and click the specific model you want OpenClaw to use to enter its details page. In this topic, "Doubao-Seed-Code" is used.

Click the copy button to obtain the "Model ID".

tip

The obtained model ID will be used to configure OpenClaw. Keep it secure.

Step 3: Configure the model service

Open Windows PowerShell ISE.

On the toolbar, click "File > New".

Copy and paste the following script content into the newly created script file.

tip

Adjust the following parameter values in the command based on your actual situation:

If you have subscribed and want to use the ModelArk Coding Plan:

--ark-coding-plan: Retain this parameter. This indicates using the ModelArk Coding Plan.

${Your_API_KEY}: Replace with ModelArk API Key. Go to "VeArk console > API keys" to create and obtain the API Key.

${Your_Model}: All replace with the name of the model supported by the Coding Plan. You can obtain the model name by referring to Configure the model name (specified in the configuration file).

If you have not subscribed to the ModelArk Coding Plan or do not want to use the Coding Plan:

We recommend that you subscribe to the ModelArk Coding Plan, which costs as low as CNY 9.9 per month and enables flexible model calls.

--ark-coding-plan: Delete this parameter. This indicates not using the ModelArk Coding Plan.

${Your_API_KEY}: Replace with ModelArk API Key. Go to "VeArk console > API keys" to create and obtain the API Key.

${Your_Model}: Replace with the ID of the model you want to use. Go to "VeArk console > Foundation models" to obtain the Model ID from the details page of the corresponding model.

${Your_Feishu_APP_ID}: Specify the App ID of the Feishu bot.

${Your_Feishu_APP_Secret}: Specify the App Secret of the Feishu bot.

openclaw onboard --skip-channels --skip-skills --non-interactive --accept-risk
openclaw plugins enable feishu

$Url = "https://openclwavolc.tos-cn-beijing.volces.com/setup-openclaw-win.py"
$Path = "$env:TEMP\setup-openclaw-win.py"

try {
Invoke-WebRequest -Uri $Url -OutFile $Path -ErrorAction Stop
Write-Host "✅ $Path"
} catch {
Write-Error "❌ $_"
exit 1
}

$PythonExe = "C:\Program Files\Cloudbase Solutions\Cloudbase-Init\Python\python.exe"

& $PythonExe $Path `
--ark-coding-plan `
--ark-api-key "${Your_API_KEY}" `
--ark-model-id "${Your_Model}" `
--feishu-app-id "${Your_Feishu_APP_ID}" `
--feishu-app-secret "${Your_Feishu_APP_Secret}"

openclaw gateway install

On the toolbar, click the "Save File" button.

Save the script to your desktop with the filename openclaw-init and click "Save".

On the toolbar, click the Run button to execute the configuration script.

If the following output is displayed, the OpenClaw service is started.

warning

During the use of OpenClaw, do not close the openclaw-gateway window. If the window is mistakenly closed, execute the openclaw gateway restart command in a new PowerShell window to restart.

Integrate OpenClaw with Feishu.

Go to Feishu Open Platform. Then, ensure that your Feishu bot has the persistent connection enabled, the required events added, and the permissions configured. For details, refer to "Configure the OpenClaw Feishu AI assistant > Steps 3-9".

You have now successfully deployed OpenClaw and integrated it with the Feishu bot. You can start conversing with the bot in Feishu to assign tasks.

Change the ModelArk model used by OpenClaw

To switch to a subscribed Coding Plan and its supported models or to another model in ModelArk "Foundation models", follow the instructions in this section to adjust the OpenClaw configuration.

Execute the following command in PowerShell to modify the OpenClaw configuration.

Use a Coding Plan
Not use a Coding Plan
tip
Replace <your_key> with the API Key you created in ModelArk.
Replace <model_name> with the model name included in the Coding Plan. You can obtain the model name by referring to Configure the model name (specified in the configuration file).

openclaw config set models.providers.ark.baseUrl https://ark.cn-beijing.volces.com/api/coding/v3
openclaw config set models.providers.ark.apiKey <your_key>
openclaw config set models.providers.ark.models[0].id <model_name>
openclaw config set models.providers.ark.models[0].name <model_name>
openclaw models set ark/<model_name>
openclaw gateway restart
tip
Replace <your_key> with the API Key you created in ModelArk.
Replace all <model_id> with the ID of the model you want to use. Go to "VeArk console > Foundation models" to obtain the Model ID from the details page of the corresponding model.

openclaw config set models.providers.ark.baseUrl https://ark.cn-beijing.volces.com/api/v3
openclaw config set models.providers.ark.apiKey <your_key>
openclaw config set models.providers.ark.models[0].id <model_id>
openclaw config set models.providers.ark.models[0].name <model_id>
openclaw models set ark/<model_id>
openclaw gateway restart

Last updated: 2026.06.18 17:48:27