agentscope-ai/agentscope-java

Build distributed, production-grade, long-running agents.

What it solves

AgentScope Java 2.0 provides a production-ready framework for building and deploying distributed, enterprise-grade AI agents. It moves beyond simple prototyping by addressing the challenges of long-running tasks, secure tool execution, multi-tenant isolation, and session recovery in production environments.

How it works

The framework uses a dual-layer architecture where a reasoning core (ReActAgent) is wrapped in a "Harness" (HarnessAgent) that adds production infrastructure. It employs a unified event stream for real-time rendering and human-in-the-loop (HITL) interactions, a middleware system for intercepting the reasoning-acting loop, and a permission system to gate tool calls. For deployment, it supports stateless horizontal scaling with external state stores (Redis, MySQL, PostgreSQL) and isolated execution environments (Docker, Kubernetes, or cloud sandboxes).

Who it’s for

Java developers and enterprise architects who need to build scalable, secure, and observable AI agent applications that can handle complex, long-term workflows and integrate into existing business systems.

Highlights

  • Harness Engineering: Includes a skill repository for self-evolution, layered memory with auto-compaction, and the ability to spawn and delegate tasks to sub-agents.
  • Enterprise Deployment: Supports multi-tenant isolation, session recovery for zero-downtime deploys, and secure sandboxing for running untrusted code.
  • Event-Driven Core: A system of 32 typed events for real-time streaming and a first-class Human-in-the-Loop (HITL) mechanism for approving sensitive actions.
  • Flexible Middleware: AOP-style hooks at six different stages (e.g., onReasoning, onActing) to extend agent behavior without modifying the core logic.

Related

  • Project
    agents-flex/agents-flex

    Agents‑Flex is a modular, Java‑first framework that bundles chat, embeddings, tool calling, RAG, vector stores, and production‑grade features (routing, observability, persistence) into interchangeable Maven modules, letting JVM developers build LLM‑driven agents, chatbots, and AI‑enhanced services without leaving the Java ecosystem.

  • Project
    Atmosphere/atmosphere

    Atmosphere is a JVM‑based framework that lets you write AI agents as annotated Java classes. It provides real‑time streaming (WebSocket, SSE, gRPC, optional HTTP/3), durable memory, policy‑driven governance, human‑in‑the‑loop approval, and adapters for many LLM providers. Agents can be exposed to browsers, Slack, Discord, MCP, A2A, etc., without rewriting code.

  • Project
    w8123/EnterpriseAgentFramework

    ReachAI is an open‑source Java‑centric platform that retrofits existing enterprise apps (OA, ERP, CRM, HR, etc.) with AI assistants. It scans UI and APIs, registers business methods as annotated Capabilities, uses LLM‑driven “AI Coding” to implement suggested refactors, and deploys Agents and versioned Workflows that can answer natural‑language queries, fill forms and trigger transactions—all while preserving original authentication, requiring user confirmation for writes, and providing full audit traces.

  • Project
    agentscope-ai/agentscope

    A production-ready agent framework that provides an SDK and a FastAPI backend to build multi-tenant, multi-session agentic applications leveraging LLM reasoning and tool use.

  • Project
    alibaba/spring-ai-alibaba

    A production-ready Java framework for building agentic and multi-agent applications, featuring a graph-based runtime for stateful workflows and a visual management platform.