Pro-code application

• Unlike low-code/no-code drag-and-drop development, pro-code applications let you build complex AI services by writing code.

• Deploy a Python project as a publicly accessible cloud-based backend API service with a single command.

• Supports two deployment modes, serverless function and K8s, to meet different needs: stateless, fast-starting applications or high-performance, stateful, long-running tasks.

• Provides one-stop MCP tool integration, letting you connect MCP services like a knowledge base, workflow, and plugin directly in the console.

• Supports a custom frontend experience with three options: Direct Experience, Custom Interactive Cards, and a Custom Frontend WebUI, letting you build rich interactive interfaces with the Spark Design framework.

• Supports enterprise-level capabilities such as automated operations, observability, log service, and api gateway.

Quick start

1. Go to the Alibaba Cloud Model Studio Application Center and click Create Application. In the Create Application dialog box, select High-Code Application on the left.

   Select a creation type:

   • Create in Console: Select a template, configure resources, and deploy with one click directly in the console. This is recommended for first-time users.

   • Create from CLI: After creating a blank application in the console, use the AgentScope-AI command-line tool to upload and deploy a local code package. This is suitable for developers who already have project code. For detailed instructions, see the API Development Guide.

   Enter an Application Name and click Confirm.

   

2. After confirming, you are taken to the application deployment configuration page. Complete the configuration in the following sections:

   During creation, follow the on-screen guidance to grant the required roles and permissions for Function Compute (FC) and api gateway. If you have insufficient permissions, contact your account administrator or IT administrator to obtain the required authorization.

   

   Select deployment mode

   • serverless function (Default): Suitable for low-load, stateless scenarios that require fast startup, with lower deployment costs.

   • K8s: Ideal for high-performance, stateful, long-running agent tasks. Before using this option, you must activate and authorize Container Service for Kubernetes (ACK).

   Submit code

   • Use template code (Default): Select a pre-configured application template to get started quickly.

   • Upload code package: Upload a locally developed code package in .whl format. To learn how to create a code package, see the API Development Guide.

   The platform provides the following application templates:

   Template	Description
   Basic conversational agent	A basic example agent focused on LLM-based conversation, suitable for quick trials and further development.
   Tool-calling agent	Includes built-in MCP tools (such as web search and a calculator) to demonstrate tool-calling capabilities.
   In-depth research agent	An in-depth research agent built on AgentScope components that can conduct internet research and generate reports.

   Select deployment resources

   Configure the instance plan (vCPU, memory, disk size), minimum number of instances, concurrency per instance, and region as needed. You can use the default settings for your first deployment.

   For latency-sensitive workloads, we recommend setting the Minimum Number of Instances to 1 or higher. This enables millisecond-level warm starts and helps keep your service available. For details, see Instance Types and Specifications.

   After finishing the configuration, click Deploy Now to begin the deployment.

   Important: Billing starts after deployment. Application testing triggers real invocations. Function Compute, api gateway, storage, and model calls all incur charges. For actual costs, see your Alibaba Cloud bill.

   Enable application tracing

   When enabled, you can analyze the call paths of your application code and monitor call parameters.

3. Wait a few minutes for the application to build and deploy. The page displays the real-time progress and elapsed time for the build and deployment. You can click the Logs link next to the progress bar to view detailed build logs and deployment logs.

   

4. After the deployment is complete, configure the Alibaba Cloud Model Studio API Key. You can then use the test panel on the right to test and debug your application. To learn how to get an API Key, see Obtain API Key.

   • Text Conversation: Interact directly with your application through conversation to quickly verify its functionality.

   • API Test: Manually call the application's API endpoints, such as GET /health and /process. This is useful for development and debugging. For details on the API protocol, see the API Development Guide.

   

Application management

After deployment, the application details page has four tabs on the left: Deployment, Tools, Gateway, and Frontend. These tabs cover the entire workflow for managing and using your application. The panel on the right offers two test modes, API Test and Text Conversation, letting you verify application functionality at any time.

Deployment

Use the Deployment tab to manage the application's deployment status, resource configuration, and runtime logs.

• Redeploy: Click the Redeploy button at the top of the page to update all deployment-related settings. This includes uploading a new code package (only .whl format is supported), adjusting application performance (by modifying resource specifications, instance count, and concurrency), and changing the region.

  Updating the configuration or restarting the service rebuilds and redeploys the application.

• Configuration: This section contains three collapsible areas: basic configuration, environment variables, and Trigger. You can view and modify the application's basic information, custom environment variables (such as an API Key), and HTTP trigger settings.

• Logs: View the application's runtime logs. You can filter logs by time range, content, and instance ID. Build process logs are available via the build logs and deployment logs links in the deployment status bar.

• To stop the application and avoid further charges, click the Stop Service button at the top of the page.

Tools

Pro-code applications support one-stop MCP tool integration. In the Tools tab, you can connect your application with MCP services such as a knowledge base, workflow, or plugin. For detailed configuration instructions, see Tool Integration.

Gateway

After your application is stable and its performance meets expectations, we recommend enabling the gateway feature. This lets you access the cloud service in a production environment using a custom domain name and gateway routes.

1. In the Gateway tab of the pro-code application details page, follow the console guide to create an Alibaba Cloud Cloud Native API Gateway based on your needs. (This will incur a small fee. For a detailed price list, see Billing Overview).

   Important: The application's region must be the same as its gateway region.

2. Create a route and configure token-based authentication. Once the new API route is active, you can access the pro-code application through the gateway.

3. Using the application in a production environment: We recommend enabling the "Prohibit Internet Access" switch for the test Trigger from being accessed over the public internet.

4. Example call:

   curl -i -X POST "http://{your-domain}/{your-agentCode}/process" \
     -H "Content-Type: application/json" \
     -H "Authorization: Bearer Replace with your application gateway authentication token" \
     -d '{
       "input": [
         {
           "role": "user",
           "content": [
             { "type": "text", "text": "Hello" }
           ]
         }
       ],
       "session_id": "xxxxxx",
       "user_id": "xxxxx"
     }'

Frontend

The Frontend tab offers three ways to create a rich interactive interface for your pro-code application:

• Direct Experience: Switch to the Text Conversation mode in the right-side panel. After configuring the conversation API, you can interact with and test the application directly.

• Custom Interactive Cards: Define custom card styles in your application code to implement interactive elements like rich text, buttons, and forms.

• Custom front-end WebUI: Build a complete front-end UI based on the AgentScope Spark Design framework. Spark Design provides two core packages: @agentscope-ai/design (UI component library) and @agentscope-ai/chat (chat component library), which support capabilities such as Markdown rendering, streaming response, and voice input. For more information, see the Spark Design online documentation.

Next steps

• Tool Integration: Learn how to connect to a knowledge base and external tool services using the MCP protocol.

• API Development Guide: Learn how to develop, upload, and deploy a pro-code application using AgentScope-AI.