Docs

Elastic Compute Service

Quickly deploying OpenClaw via application templates

Quickly deploying OpenClaw and integrating DingTalk AI assistant

OpenClaw is an open-source, local-first AI agent and automation platform. With OpenClaw, you can deeply integrate multi-channel communication capabilities with large language models to build a customized AI assistant with persistent memory and proactive execution capabilities.

Change history

warning

You cannot integrate the DingTalk application with OpenClaw deployed on an ECS instance purchased before 24:00 on February 2, 2026. Reinstall the instance operating system by creating an OpenClaw application update task, and then perform the integration by referring to "Configure the model and integrate the DingTalk application after ECS instance purchase".

To upgrade or reset OpenClaw deployed on the ECS instance, you can create an OpenClaw application update task to complete the operation.

Update time	Description
March 5, 2026	Added support for the latest features of the Community Edition released on March 2, 2026, as well as compatibility with the Feishu plugin released on March 4, 2026.
February 27, 2026	Added support for the latest features of the Community Edition released on February 23, 2026.
February 12, 2026	Integrated the AI Assistant Security feature. Once the feature is enabled, enterprise-grade security protection is provided for OpenClaw.
February 10, 2026	Fixed the conflict issue with the Feishu plugin and aligned with the latest features of the Community Edition up to February 6, 2026.
February 2, 2026	Added support for integration with DingTalk applications.

Features

Volcano Engine provides an out-of-the-box application template for rapid OpenClaw deployment. This template not only helps you quickly deploy OpenClaw and integrate it with a DingTalk application, but also comes with pre-integrated common Skills from AgentKit. In addition, it natively supports multimodal capabilities for image and video generation.

For developers, especially for individual development scenarios, ModelArk offers a Coding Plan solution that supports the use and free switching of mainstream code models (the latest versions of Doubao-Seed-Code, GLM, Kimi, DeepSeek, and more). This facilitates coding tasks such as personal project development, learning and practice, and tool building.

For enterprise-level development needs, you can select any large language model from "Foundation models" of ModelArk. Multiple model services, including Doubao-Seed-Code, GLM-4.7, DeepSeek-V3.2, and Kimi-K2-Thinking are now fully compatible with the OpenClaw technical ecosystem.

warning

To use the natively supported multimodal capabilities for image and video generation, go to the Model activation page, and then select either doubao-seedream-4-5 (251128) or doubao-seedance-1-5-pro (251215) for OpenClaw.

Prerequisites

ModelArk has been activated, and the API Key has been created and obtained.

Procedure

Step 1: Create a DingTalk application

Create a DingTalk application.

tip

To create a DingTalk application, your DingTalk account must have developer permissions. You can contact your organization administrator to obtain developer permissions for the DingTalk Open Platform.

In the directory tree on the left, select "Enterprise internal applications > DingTalk applications", and click the "Create application" button in the upper right corner of the page.

Configure the name, icon, and other information of the application, and click the "Save" button.

Grant the application permission to send card messages.

Log in to the DingTalk application console, and click the name of the application you created earlier to enter its details page.

In the directory tree on the left, select "Development configuration > Permission management".

Enter Card in the search box, select "Interactive card instance write permission" and "AI card streaming update permission", and then click the "Batch apply" button to complete the operation.

Enter qyapi_robot_sendmsg in the search box, and click the "Activate now" button in the operation column to the right of "Intra-enterprise bot message sending permission" to complete the operation.

Configure the DingTalk bot.

In the directory tree on the left, select "Application capability > Add application capability", and click the "Add" button on the "Bot" card.

On the bot configuration page, turn on the bot configuration switch.

Set "Message receiving mode" to "Stream mode".

Click the "Publish" button to save the configuration.

In the pop-up window, click the "Publish" button to complete the editing.

tip

Keep "Bot configuration" enabled after publication.

Obtain the application information.

In the directory tree on the left, select "Basic Info > Credentials & Basic Info".

Copy the Client ID and Client Secret.

tip

The obtained Client ID and Client Secret will be used to configure OpenClaw. Keep them secure.

Publish the application version.

In the directory tree on the left, select "Application publication > Version management and release", and click the "Create new version" button.

Configure the version number, visibility scope, and other information of the application as needed, and click the "Save" button.

In the pop-up window, click the "Confirm publication" button to publish the application version.

Step 2: Create an ECS instance

Click here to log in to the ECS console.

Quickly purchase an ECS instance pre-installed with OpenClaw by using an application template, and configure the following parameters.

Parameter	Description
Instance configuration
Application template	Select "OpenClaw".
Billing method	Select the billing method of the instance. In this topic, "Subscription" is selected as an example. Learn more.
Region	Select the region where the instance is located. In this topic, China (Shanghai) is used as an example. tip Currently, only China (Guangzhou), China (Beijing), China (Shanghai), and Asia Pacific (Jakarta) are supported. Learn more.
Package specification	Select the instance type. Select 2 vCPU, 2 GiB memory or higher. This topic uses the "e-c1m1, 2 vCPU, 2 GiB" type as an example.
Login password	Configure the instance login password.
Confirm password	Re-enter the instance login password for confirmation.
Security protection	Select whether to enable the AI Assistant Security feature. (Recommended) Enable: Read the AI Assistant Security SDK Protocol and check to enable the AI Assistant Security feature. tip Once enabled, the AI Assistant Security feature monitors and defends against three core risks in real time without affecting the normal operation of OpenClaw: prompt injection, PII leakage, and risky operations. Not enable: Do not use the related services.
Role authorized	Select whether to complete the required role authorization for the cloud assistant. (Recommended) Authorize: Check the "Confirm authorized Cloud assistant roles" option. You can perform version upgrades, data synchronization, and other related operations only after the authorization is complete. tip After authorization, you can use the application management feature to view your Coding Plan quota usage, upgrade OpenClaw versions, and manage its associated models and message channels. Learn more. Do not authorize: You cannot use the related services.
Advanced configuration
Model configuration	Select and configure a model on the ModelArk platform. OpenClaw will use this model for inference after your ECS instance starts. Configuration method: Foundation models: Supports LLMs from "Foundation models" of ModelArk. Multiple model services, including Doubao-Seed-Code, GLM-4.7, DeepSeek-V3.2, and Kimi-K2-Thinking, are now fully compatible with the OpenClaw technical ecosystem. tip If you have not activated the ModelArk service, after you select a model and create an instance, the platform will automatically activate the selected model service for you. After that, the model will be billed by token. OpenClaw deployed via the "OpenClaw" application template is integrated with common Skills from AgentKit. It natively supports multimodal capabilities like image and video generation. To use related capabilities, first enable the corresponding models, and select the target models such as doubao-seedream-4-5 (251128) or doubao-seedance-1-5-pro (251215). Coding Plan: ModelArk Coding Plan is a subscription service tailored for developers for AI coding scenarios. It supports the use and free switching of mainstream code models (including the latest versions of Doubao-Seed-Code, GLM, Kimi, DeepSeek, and more). tip You can click the "Subscribe now" button to go to the ModelArk Coding Plan campaign page and subscribe to a package as needed. Model selection: Select the model that provides inference services for OpenClaw. API Key: Select the API Key you want to use from the drop-down list. warning Keep your ModelArk API Key secure. If leaked, it could allow third parties to impersonate you for unauthorized use, causing unexpected token fees.
Message channel integration	Select DingTalk. After the ECS instance starts, OpenClaw will automatically use DingTalk as the message notification channel. Client ID: The client ID of the target DingTalk application. Client Secret: The Client Secret of the target DingTalk application. tip You can log in to the DingTalk application console and obtain the Client ID and Client Secret in the "Basic Info > Credentials & Basic Info" section of the target application details page. Caution: The Client ID and Client Secret are important identity credentials for accessing the DingTalk application. Keep them secure to prevent disclosure.

In the "Configuration details" section on the right, confirm the instance configuration and fee, and check "I have read and agree with Elastic Compute Service Terms of Service".

Click the "Buy now" button to complete the purchase.

You can also click the "Go to console" button to view the specifications of the newly created instance on the instance list page. You can also log in to the instance through the console.

Step 3: Check service startup status

Execute the following command to check the running status of the OpenClaw service and check whether it started as expected.

systemctl --user status openclaw-gateway.service

If the service status in the output is active (running), the OpenClaw service has started as expected. You can refer to the subsequent steps to use the AI assistant in DingTalk.

Use the AI assistant

After the configuration is complete, you can have a private chat with the AI assistant on DingTalk to assign tasks, or add it to a group chat and mention it with @ to assign tasks. This section uses the scenario of adding it to a group chat to interact with it as an example.

Enter the "Group settings" page in the DingTalk group.

In the "Management" module, click the bot card.

On the bot management page, click "Add bot".

Search for and select the DingTalk application you created.

Click the "Add" button, confirm the information on the "Add bot" tab, and click the "Complete" button to complete the operation.

Mention the bot with @ in a DingTalk group chat to interact with it directly.

The following two example tasks are used to describe how the AI assistant works.

Task 1: Generate Spring Festival-themed images

Prompt used:

Spring Festival theme. Use the image_generate tool to create four storyboard images in sequence. Create four independent tasks at once. If no style is specified, use "Shanghai fine arts cartoon style" by default. Send me all four images after generation.

Task 2: Create a video from the images generated in "Task 1"

Prompt used:

Use these 4 images (with the first and last frames retained) to generate 3 video clips with voiceover. Use the video-generate tool with 480P resolution configuration. The size of a single file should not exceed 5 MB. Compress the generated MP4 files and send them to me.

Last updated: 2026.09.08 14:33:21