DocsICP FilingConsole
Documentation
Home Page Elastic Compute Service User Guide Elastic Compute Agent Getting started

Getting started

Updated at:
Copy as MD
Product
My Favorites

This quickstart walks you through a console session with Elastic Compute Agent to experience its four core capabilities: AI chat, resource query, anomaly diagnosis, and health inspection. Optionally, learn how to set up scheduled tasks for automated O&M checks and view Credits usage.

Prerequisites

  • An Alibaba Cloud account with identity verification completed. When you first access the Elastic Compute Agent console, select the Terms of Service checkbox and click Activate Now to enable the service. No subscription or upfront payment is required.

  • If you want to use Elastic Compute Agent as a RAM user, the Alibaba Cloud account must first grant the required permissions. For more information, see Permission Configuration.

  • (Recommended) At least one ECS instance in your account to experience resource query, health inspection, and other capabilities.

Start your first session

The home page displays entry cards for the four core capabilities: AI chat, resource query, anomaly diagnosis, and health inspection. Click any preset question to get started.

Example 1: AI chat

  1. On the home page, click the preset question "What is the difference between pay-as-you-go and subscription billing for ECS?"

  2. The Agent searches the knowledge base and returns a structured comparison table.

  3. Expected consumption: 2 Credits.

Example 2: Resource query

  1. On the home page, click the preset question "Query ECS instance information in Hangzhou region," or type a natural language query in the input box.

  2. The Agent calls the resource query tool and returns an instance list with instance ID, instance type, status, and region.

  3. Expected consumption: 5–10 Credits (depending on the number of instances).

Example 3: Anomaly diagnosis

  1. On the home page, click the preset question "Help me diagnose the ECS instance that cannot be connected remotely."

  2. The preset question is automatically filled in the input box. Add the specific instance ID (you can find instance IDs in the result of Example 2) and click Send.

  3. The Agent delegates a sub-agent for diagnosis and returns a structured diagnostic report with issue location and fix suggestions.

  4. Expected consumption: 10–20 Credits (single instance).

Example 4: Health inspection

  1. On the home page, click the preset question "Inspect GPU instances in the current account."

  2. The Agent calls the inspection tool to scan all GPU instances for running status, GPU hardware health, driver version, and monitoring metrics. It then returns a structured inspection report covering inspected instances, anomalies, and fix suggestions.

  3. Expected consumption: 20–50 Credits (depending on the number of instances; single one-time inspection).

Note

For periodic automated health inspections (for example, daily scheduled execution with results pushed to an IM group), use the Scheduled Tasks feature. See Create a scheduled task (optional) below.

Create a scheduled task (optional)

To periodically run automated O&M checks (for example, daily GPU instance health inspections), create a scheduled task based on an official template. The following example shows the basic configuration process without IM push notifications:

  1. In the left-side navigation pane of the Elastic Compute Agent console, click Scheduled Tasks, and then click the Task Templates tab.

  2. Select a built-in template (for example, GPU Resource Health Check) and click Create Task.

  3. In the task form, enter the task name and select the execution schedule (for example, "09:00 every day"). Keep the Enable IM Channel toggle disabled so that execution results are only viewable in the console and are not pushed to an IM group.

  4. After saving the task, click the Configured Tasks tab. The task card displays "Triggered automatically at 09:00 every day." Click Execute Now to manually trigger a test run.

  5. Click the Execution Records tab to view historical execution results and detailed reports.

Note

To automatically push scheduled task results to DingTalk, Lark, or WeCom groups, create an IM channel first and enable the Enable IM Channel toggle in the task form. For more information, see Configure IM Channel.

View Credits usage

In the lower-left corner of the Elastic Compute Agent console, click Settings > Usage and Limits to view the total Credits, consumed Credits, remaining quota, and consumption trends for the current account. For detailed field descriptions and billing cycle information, see Billing Method.

Next steps

Previous：What is Elastic Compute Agent?Next：Billing