Learn about Flow Agent application types, extended capabilities, and how to get started.
What is Flow Agent
Flow Agent is a visual application development platform from Alibaba Cloud Model Studio. It supports building, debugging, and publishing agent and workflow applications.
Application types
Large language models have limitations in processing private knowledge, accessing real-time information, following fixed procedures, and planning complex tasks. Model Studio integrates knowledge base retrieval, external tool calls, and memory to help you quickly build AI applications that solve business problems.
Choose by task: use an agent application for open-ended conversations or answers based on private documents; use a workflow application to generate a report through steps you have already defined.
| Application type | Description | Best for |
|---|---|---|
| Agent application | Natural language configuration (no-code); calls tools such as knowledge bases and MCP services to complete tasks | Customer support, knowledge Q&A, task assistance |
| Workflow application | Visual node orchestration (low-code); chains multi-step tasks into stable and repeatable execution paths | Report generation, order processing, multi-step approvals |
Extended capabilities
| Capability | Description | Best for |
|---|---|---|
| File Q&A | Understand and answer questions about documents, images, audio, and video, with full-text citation, chunk retrieval, and custom processing modes | Long-document Q&A, research, and understanding image and audio/video content |
| Prompt | Manage reusable Prompt templates with a sample library, automatic optimization, and feedback-based optimization | Reusing common prompts and continuously improving response quality |
| Plugins | Extend application capabilities with official and custom plugins | Web search, code execution, and business system integration |
| Task Center | View background tasks from asynchronous calls, including status, input/output, and timing | Tracking long-running tasks and their results |
Get started
Create and configure your chosen application, then use its API to integrate it into your system. For missing workflow results or unexpected variables, inspect node inputs and outputs in Test application. For call timeouts, see Call via API. If an API returns an error code, see Error codes.
For managed sessions, isolated sandboxes, and long-running tasks, see the separate Managed Agents overview.
Quick start
Create your first agent application in 5 minutes.
Create an agent application
Select a model and write a prompt to create an agent application.
Create a workflow application
Orchestrate nodes visually to create a workflow application.
Use the CLI
List and invoke published applications and workflows from the terminal with bl app.
Practical examples
- Agent applications: Create an AI ecommerce customer service assistant for Q&A and Build a travel planning agent with Amap MCP.
- Workflow applications: Create an AI ecommerce customer service assistant for complex processes and Build a low-code data labeling workflow.
API integration
- New agent applications (Agent 2.0, recommended): New agent application API.
- Workflow applications: Call workflow applications.