Agentic AI is an open source framework for building agents, graph workflows, and multi-agent applications in Java. It was forked from Spring AI Alibaba and focuses on stateful agent runtime capabilities: graph orchestration, persistence, context engineering, human-in-the-loop workflows, and embeddable debugging tools.
Agentic AI also supports multi-model integration capabilities provided by Spring AI, with optional extensions for model providers, MCP, A2A, Nacos, storage, and tool execution.
- argi: core runtime, Agent Framework, Graph Core, Spring Boot starters, and embedded Studio.
- argi-extensions: Optional integrations for model providers, A2A, Nacos, storage, tool execution, and graph nodes.
- argi-example: Runnable samples for chatbots, agents, graph workflows, and integrations.
- argi-website: ARGI website.
- Connect AI protocol conversion across OpenAI Chat, Responses, Anthropic, and Gemini.
<groupId>io.github.agentic-ai</groupId>The core Java package prefix is io.github.agentic.ai.
Agentic AI is independently maintained by the community. It originated from Spring AI Alibaba community work but is not an official Alibaba project.
The projects are released under the Apache License 2.0. Issues, discussions, and contributions are welcome.