AgentScope Java
ActiveDescription
Agent-oriented programming framework for building LLM applications in Java. Provides agent abstractions, tool calling, multi-agent collaboration, and other core capabilities for enterprise Java ecosystem integration.
Key Features
- ReAct (Reasoning-Acting) paradigm for autonomous task planning and dynamic tool selection
- Production-grade runtime intervention: safe interruption, graceful cancellation, and human-in-the-loop hooks
- Built-in tools including PlanNotebook task decomposition, structured output parsing, and long-term memory with semantic search
- MCP and A2A protocol support for extending capabilities and distributed multi-agent collaboration
- High-performance reactive architecture with GraalVM native image support achieving 200ms cold start
- Security sandbox for untrusted tool code with pre-built environments for GUI, file system, and mobile interactions
Use Cases
π‘ Building enterprise-grade LLM-powered applications in Java ecosystems
π‘ Multi-agent collaboration systems with service discovery via A2A protocol
π‘ Production deployments requiring runtime intervention and human oversight
π‘ Serverless and auto-scaling environments benefiting from GraalVM native compilation
π‘ RAG systems integrating enterprise knowledge bases with agent reasoning
Tags
Categories
Quick Start
1. Add Maven dependency (JDK 17+):
```xml
<dependency>
<groupId>io.agentscope</groupId>
<artifactId>agentscope</artifactId>
<version>1.0.12</version>
</dependency>
```
2. Create a ReAct agent:
```java
ReActAgent agent = ReActAgent.builder()
.name("Assistant")
.sysPrompt("You are a helpful AI assistant.")
.model(DashScopeChatModel.builder()
.apiKey(System.getenv("DASHSCOPE_API_KEY"))
.modelName("qwen-max")
.build())
.build();
```
3. Call the agent: `Msg response = agent.call(Msg.builder().textContent("Hello!").build()).block();`
4. See docs: https://java.agentscope.io/