Agent application

Large language models cannot access proprietary knowledge bases or retrieve real-time information. Model Studio agent applications let you connect an LLM to external tools and knowledge sources without code, extending the model beyond its built-in capabilities. Important

• Console restrictions: Only International Edition users who created applications before April 21, 2025 can access the Application Development tab, as shown in the following figure.

  This tab contains the following features: Applications ( agent application and workflow application), Components ( prompt engineering and plug-in), and Data ( knowledge base and application data). These are all preview features. Use them with caution in production environments.

• API call limits: Only International Edition users who created applications before April 21, 2025 , can call the application data, knowledge base, and prompt engineering APIs.

How it works

An agent is driven by a prompt and completes complex tasks by orchestrating external capabilities. After receiving a request, the LLM identifies the user's intent and plans the necessary tasks. The agent then autonomously calls external tools and integrates the results into a final response.

Use an agent when you want the model to autonomously decide how to approach a task using available tools --- for example, building a customer support assistant that queries a private knowledge base, calls a live data API, and summarizes findings in a single response.

Model Studio agents support these capabilities:

1. Knowledge base (RAG): Connects the agent to an external knowledge base to answer questions based on private data. Use this when your agent needs to answer domain-specific questions that the model's built-in knowledge cannot cover.

2. Plugin: Calls pre-built platform tools, such as code execution, image generation, or weather queries. Use this when the agent needs to perform specific actions beyond conversation.

Quick start

Create a basic agent

1. Go to the Application Management page in the Alibaba Cloud Model Studio console. Click + Create Application . On the Agent Application tab, click Create Now.

2. On the Application Configuration page, select a model from the model selection drop-down menu, such as Qwen-Plus. You can use the default settings for other parameters.

3. After creation, enter Hello in the dialog box on the left to test.

Agent capabilities

Extend an agent by selecting models, optimizing the system prompt, adding a knowledge base (RAG), and calling plugins.

Model

The model drives the agent's reasoning and decision-making. Model Studio agents support Qwen series models and custom-deployed models.

1. Select a model

   From the model selection drop-down menu, select a model, such as Qwen-Plus. Click More Models to select other models.

2. Configure parameters

   Click to the right of the model drop-down list to configure parameters:

   1. Maximum response length: The length limit for the model's generated response, excluding the prompt. Varies by model.

   2. Context turns: The maximum number of past conversation turns included in the model's input. More turns increase conversational relevance.

   3. temperature: Controls the randomness of generated responses. Higher values increase diversity; lower values increase consistency. Range: \0, 2).

   4. enable_thinking: Enables thinking mode. Unavailable for models that do not support thinking mode.

      When thinking mode is enabled, the model performs deeper reasoning before responding, which increases token consumption.

System prompt

A system prompt defines the agent's role, behavior, and constraints, keeping responses consistent and task-oriented. Write an effective system prompt by:

• Defining a persona: Describe who the model should act as and what expertise it should draw on.

• Specifying output format: Describe the structure, length, or style you expect.

• Setting constraints: Tell the model what to avoid or what rules to follow.

• Guiding tool use: Mention specific tools by name and describe when they should be used.

  1. Configure a prompt

     Set the system prompt to Please answer my question in the style of 'One Hundred Years of Solitude'. Compare the results:

     • Without a system prompt:

     • With the system prompt:

Knowledge base (RAG)

Retrieval-augmented generation (RAG) lets an agent query external knowledge bases and use the retrieved information to generate answers. For private or domain-specific questions --- such as internal product documentation or proprietary datasets --- RAG improves accuracy and reduces hallucinations. For more information, see Knowledge base.

Note: Retrieved text consumes the model's context window. Adjust the retrieval strategy and text length to balance coverage and context window limits.

Plugin

An agent can call plugins for specific tasks, such as code execution, web searches, or text-to-image generation. Use plugins when the agent needs to take actions or generate content that goes beyond its built-in knowledge. Model Studio provides official plugins and supports custom plugins. For more information, see Plugin overview.

Agent interaction

Text conversation

Text conversation is the primary interaction method and supports multi-turn conversations.

Text conversation supports two input methods:

1. Text input: Enter text to converse with the agent.

2. File upload: Upload files such as documents, images, videos, and audio as attachments for the agent.

Publish and call an agent

Model Studio agents can be called via API. They also support one-click publishing to third-party platforms and can be integrated into business processes as components or through the Moflow channel.

Publish your application

Important

Publish your application before calling or integrating it.

Click Publish in the upper-right corner of the agent application management page, and then click Confirm Publish. If you have published before, a pop-up shows changes since the last publication. Note

If a RAM user created the application, they must have the ram:CreateServiceLinkedRole permission for the Service-Linked Role before publishing. For more information, see Service-Linked Role.

API call

On the Publish Channel tab of the agent application, click View API next to API Call to view the API call methods.

Replace YOUR_API_KEY with your actual Model Studio API key before you invoke the API.

Agent management

Copy and delete

On the My Applications page, find the application card. From the More > Copy Application/Delete Application menu, you can copy, delete, or rename the agent.

When to copy an application:

• Creating test versions with different prompts or models

• Adapting an agent for different audiences or use cases

• Creating a backup before making major configuration changes

Version management

Version management lets you edit version descriptions or revert to a previously published version.

1. On the Configure tab of the agent application, click Version Management in the upper-right corner of the top navigation bar.

2. In the list of historical versions, select a version.

   • To edit version details, click the icon. In the Edit Version Description dialog box, make your changes and click OK.

   • To revert to this version, click Overwrite Current Draft , and then click Confirm.

     This version overwrites the current draft.

Billing

Agent billing includes:

1. Model calls

   Agents incur model call fees based on model type and token usage.

   For billing details, see the [Model Studio console.

2. Knowledge base

   • The knowledge base feature is free for a limited time.

   • Text chunks retrieved from the knowledge base increase input tokens, which may increase model call fees.

3. MCP

   • Some official MCPs are billed based on model calls, such as text-to-image, text-to-video, and speech synthesis MCPs.

   • Some MCP services involve third-party API calls, which may incur fees. These fees are charged by the third-party service provider. Model Studio does not charge for these calls.

4. Long-term memory

   • Data storage for long-term memory is free.

   • During Q&A, memory content is merged into the prompt and sent to the model, increasing token consumption. Memory tokens are not currently billed.

Supported models

Note

Data updates may be delayed. The agent application displays the current list.

• Qwen-Plus

• Qwen-Max

• QwenVL-Max

• QwenVL-Plus

• Qwen-Turbo

FAQ

How are Model Studio applications billed? Creating an application is free. Q&A calls incur model call fees based on the model type. I configured a knowledge base, but the agent's answers are not related to the knowledge base content. How do I fix this? Run a knowledge base hit test to check the similarity score between the question and the knowledge base content. If the score is low, adjust the retrieval configuration so the model prioritizes knowledge base results.

In the prompt settings, instruct the model to answer only from knowledge base content, not its built-in knowledge. If the issue persists, try a different model for more stable output. Is there a timeout limit for custom plugins? Yes, 5 seconds. Can I create agent applications by using an API? You can use the Assistant API to create applications similar to agent applications. However, Assistant API applications cannot be managed in the console . For more information, see the Assistant API documentation.