Skip to main content

Overview

High-Code Applications are built on AgentScope and AgentScope-Runtime, letting you build modular Agents in full Python code with cloud deployment and operations—suited for complex scenarios that require fine-grained control.

High-Code Applications (Pro-Code Applications) are built on AgentScope and AgentScope-Runtime. You write modular Agents in full Python code, with an end-to-end loop covering flexible construction, one-click deployment, and online operations—suited for complex scenarios and long-running tasks that require fine-grained control over logic details.

Comparison with Other Application Types

DimensionAgent ApplicationWorkflow ApplicationHigh-Code Application
DevelopmentZero-code, console configurationLow-code, visual orchestrationFull-code, Python development
Logic controlPrompts and tool configurationNode orchestration and branchingComplete code logic, any complexity
RuntimeShared runtimeShared runtimeDedicated deployment (Serverless Function or K8s)
Best forQ&A, chat, lightweight tasksProcess-oriented, reusable SOPsDeep customization, complex business logic, integration with existing systems
OperationsPlatform-managedPlatform-managedLogs, observability, gateway, custom frontend

Core Concepts

ConceptDescription
ApplicationA deployed High-Code Agent service, packaged as a .whl file with a main.py entry and a GET /health health-check endpoint.
Deployment modeThe runtime carrier: Serverless Function (default—low load, stateless, fast startup) or K8s (high performance, stateful, long-running tasks; requires ACK).
TriggerThe HTTP entry of the application; the test domain comes with a trigger. For production, gateway routing is recommended.
GatewayThe Cloud-Native API Gateway, providing custom domains, routing, and token authentication for production traffic.
ObservabilityReports call paths via the @trace decorator, tracking call volume, token consumption, and latency.
Frontend experienceThree options: direct interaction, custom interaction cards, and custom WebUI based on Spark Design.

Creation Methods

High-Code Applications are created and managed on the console's Sandbox > High-Code page (https://agent.console.aliyun.com/sandbox/high-code), not in the Application Center. Two creation methods are supported:
  • Console creation: Upload code packages quickly through a visual interface—suitable for prototype validation. Choose a preset template (basic chat Agent, tool-calling Agent, or deep research Agent) or upload a locally developed .whl package. See the Quickstart.
  • CLI creation: Upload and update code packages using the runtime-fc-deploy command-line tool—suitable for automated production. See Develop & Deploy.
High-Code Applications run on Function Compute (FC) and API Gateway. On first creation, follow the console guide to complete role authorization for both services; if you choose the K8s deployment mode, enable the ACK container service in advance. Deployed applications incur a small fee based on deployment duration (less than 0.1 CNY per hour).
  • Quickstart: Create and deploy your first High-Code Application from the console.
  • Develop & Deploy: Local development conventions, observability instrumentation, CLI upload and updates.
  • MCP Tool Integration: Associate knowledge bases, MCP services, and application components.
  • Application Management: Managing the Deployment, Tools, Gateway, and Frontend tabs.
  • Best Practices: Project structure, tool selection, and debugging tips.