Use OpenClaw to integrate multi-channel communication with a large language model and create custom AI agents with persistent memory that can proactively execute tasks.
OpenClaw is an open-source AI assistant. Before you use it, fully assess its security and stability and strictly follow the license agreement to ensure the security of your system environment and data.
Billing
Deploying and using the OpenClaw service on a Simple Application Server instance incurs the following fees:
Simple Application Server fees: For more information, see Simple Application Server billing overview.
OpenClaw invokes models from Model Studio by default. These invocations are billed in two primary ways:
Coding Plan (recommended): Uses a fixed monthly fee model and provides a monthly request quota. Calls that exceed the time-based limits return an error and do not incur charges. This helps you avoid unexpected costs.
The Coding Plan currently supports models such as
qwen3.5-plus,kimi-k2.5,MiniMax-M2.5, andglm-5. For a detailed list of models, see Coding Plan overview.Billed by token usage: For
OpenClaw 2026.2.26and later versions, the default model for text and image processing isqwen3.5-plus. For billing details, see Qwen Plus.For default models and billing details of other versions, see Default models for other versions.
ImportantEnable Stop on Free Quota Exhaustion (Singapore region only): When you first activate Alibaba Cloud Model Studio (in the Singapore region), a free quota for new users is automatically issued for each model. Once the free quota is exhausted, billing is based on the prices and usage shown in the model list. To avoid unexpected charges, we recommend enabling the Stop on Free Quota Exhaustion feature. For more billing information, see OpenClaw (formerly MoltBot, Clawbot) purchasing and billing FAQ.
Purchase an OpenClaw package
If you already have a Simple Application Server instance, you can skip this step and deploy OpenClaw by resetting the system. When you reset the system, select the OpenClaw image under application images. However, resetting the system erases all data on the system disk, so back up your data first. For instructions, see Reset system.
Purchase a Simple Application Server instance with the OpenClaw application pre-installed.
On the Simple Application Server page, resource configurations are preset. To change these settings, review the options below.
Instance: The instance must have at least 2 GiB of memory. You can switch packages based on your needs.
Region: The default is US (Virginia).
Purchase Configuration: The default subscription period is 12 months. You can change the duration as needed.
Configure OpenClaw
In the Simple Application Server console, click the instance ID on the server card. On the Server Overview page, click the Application Details tab.
In the OpenClaw usage steps area, click Execute Command under Port Allow to open the firewall for the OpenClaw service's running port.
ImportantTo prevent malicious scanning and targeted attacks, OpenClaw no longer uses a fixed default port during initialization. Instead, it automatically generates a random port. You can view the OpenClaw port number in the console.
Opening the port exposes the service to the public network. You can disable public network access to the WebUI with one click. For details, see How to enable or disable public network access for the OpenClaw WebUI?
Click Execute Command under Configure OpenClaw to configure the Model Studio API key.
Two types of Model Studio API keys are currently supported:
Coding Plan exclusive API key (recommended): Uses a fixed monthly fee model and provides a monthly request quota. Calls that exceed the time-based limits return an error and do not incur charges. This helps you avoid unexpected costs.
The Coding Plan currently supports models such as
qwen3.5-plus,kimi-k2.5,MiniMax-M2.5, andglm-5. For a detailed list of models, see Coding Plan overview.A Model Studio API key billed by token usage.
You can configure the API key using either System Recommended or Manual Input. The System Recommended option provides the API key for the Model Studio Coding Plan (for cost control) and the API key for the Model Studio service closest to your server (for lower latency). Select Manual Input to use an API key from a different region or account.
System Recommended (dropdown selection)
After making your selection, click Next.
Instance region
Recommended API key region
Coding Plan API key region
Chinese mainland regions
China (Beijing)
Singapore
US and European regions
US (Virginia)
China (Hong Kong) and other Asian regions
Singapore
Manual Input
Click the button to switch to manual input. Enter the Model Studio API key, select the corresponding region for the API key, and click Next.
ImportantWhen configuring manually, ensure that the selected region for the API key is correct. Otherwise, model invocations will fail.
Click Get Address under Access Web UI Panel to get the address for the OpenClaw WebUI, where you can interact with the OpenClaw Agent on the web page.
As needed, you can refer to Disable public network access for the OpenClaw WebUI to disable public access to the WebUI with a single click.
(Optional) View the token.
In , click View to retrieve and save the token.
WarningDo not expose the full URL containing the token. This link includes authentication credentials. Anyone with this link can bypass login verification and gain administrator access to the OpenClaw console.
Configure messaging channels
OpenClaw can communicate with you through various chat applications. All channels support text messages, but support for replies with images, videos, and other media may vary by channel.
OpenClaw personalization
Generic agents may not be suitable for specific business contexts. OpenClaw provides personalization capabilities to tailor its personality and behavior for specific scenarios, such as for immersive role-playing or as a formal enterprise customer service agent. For configuration examples, see OpenClaw personalization templates and scenario examples.
The configuration files include the following:
Personality definition (SOUL.md): Defines the AI's core values, conversation style (such as formal, humorous, or concise), and code of conduct.
Identity definition (IDENTITY.md): Defines the AI's name, self-awareness, background, and role.
Workflow definition (AGENTS.md): Defines the logic for how the AI handles tasks, the rules for using tools, and the workflow.
Configuration steps: In , click Edit to modify the different configuration files. For scenario-specific configuration examples, see OpenClaw personalization templates and scenario examples.
FAQ
How to restart the OpenClaw gateway?
If you encounter connection interruptions or service unavailability, you can restart the OpenClaw gateway from the console. In , click Execute Command.
How to view the OpenClaw port number?
To prevent malicious scanning and targeted attacks, OpenClaw no longer uses a fixed default port during initialization. Instead, it automatically generates a random port. In , click Execute Command to view the OpenClaw port number.
How to change the OpenClaw model?
OpenClaw is integrated with Alibaba Cloud Model Studio, allowing you to switch between different models on the page. In , delete the default model, and then select a different Model Studio model from the dropdown list.
You can also manually enter a model name. Model codes can be found on the Model Studio Marketplace page.
How to manage public WebUI access?
You can enable or disable public network access to the OpenClaw WebUI with a single click based on your needs. In , click Enable/Disable to control public network access to the OpenClaw WebUI.
Why is my API key not recognized?
The system automatically detects the nearest Model Studio endpoint based on your server's region and displays the corresponding API key. If you are subscribed to a Coding Plan, its API key is also displayed. If no API key for a nearby endpoint exists, you can create an API key or subscribe to a Coding Plan in the corresponding region as prompted by the interface.
Instance region | Recommended API key region | Coding Plan API key region |
Chinese mainland regions | China (Beijing) | Singapore |
US and European regions | US (Virginia) | |
China (Hong Kong) and other Asian regions | Singapore |
Why is my OpenClaw's default model not qwen3.5-plus?
Currently, only the latest image version, OpenClaw 2026.2.26, uses the qwen3.5-plus model by default. The default models for other versions are as follows:
For versions
OpenClaw 2026.2.3andOpenClaw 2026.2.9,qwen3-max-2026-01-23is the default text processing model (Primary Model), andqwen3-vl-plusis the image processing model (Image Model). For billing information, see Qwen Max and Qwen-VL.In the
OpenClaw(Moltbot) 2026.1.27-beta.1version, the default model for servers purchased on or after January 31, 2026 is theqwen3-max-2026-01-23model. For servers purchased before January 31, 2026, the default model is theqwen-vl-plusmodel.
You can manually change the model that OpenClaw calls.