Spring AI Alibaba Enters Maintenance Mode as Alibaba Prioritizes AgentScope

Relationship: Upstream and Downstream

Spring AI is the official Spring community project, positioned as the foundational framework for AI application development , providing model integration, tool calling, vector storage, and other "atomic abstractions". It can be understood as the Java equivalent of LangChain.

Spring AI Alibaba was originally built on top of Spring AI , jointly maintained by Alibaba and the Spring community, positioned as an Agent intelligent agent development framework . It adds Graph workflow orchestration, multi-agent collaboration, and other capabilities on top of Spring AI — effectively the Java equivalent of LangGraph.

Thus they are not competitors but an upstream/downstream relationship : Spring AI Alibaba depends on Spring AI for base capabilities.

The "Stop Maintenance" Truth: Heart Transplant, Not Abandonment

The rumor of "stopped maintenance" is half true, half misinterpretation .

True part: Spring AI Alibaba's last formal release v1.1.2.2 was published on March 10, 2026 , and by September there had been a full half-year without a new formal release. For a project with 10k+ stars, this is indeed unusual.

Misinterpreted part: The project is not "dead". GitHub commit records show that from July to August the main branch merged 137 commits , with the latest commit on August 25 . However, these commits are mainly fixes, tests, and documentation — no major feature development — meaning the project is in "maintenance" mode, not "construction" mode.

Real Signal: Resources Tilting Toward AgentScope

More telling is the internal Alibaba comparison :

• Spring AI Alibaba : formal release every half year.

• AgentScope Java : 2.0.2 (Sept 3) → 2.0.3 (Sept 7) , a new version every four days .

The official blog foreshadowed this divergence a year ago: Spring AI Alibaba centers on Graph (workflow orchestration) , while AgentScope centers on Agentic (model-driven autonomy) , evolving independently. The last formal release v1.1.2.2's biggest new feature was first integration of AgentScope Java — wrapping AgentScope's ReActAgent into its own Graph workflow. This looks like paving the way for a handover : Alibaba bets the future of "model-driven autonomy" on AgentScope, while Spring AI Alibaba retreats to the "workflow orchestration" position and continues aligning with upstream Spring AI.

What This Means for Your Technology Selection

If you are evaluating Java AI frameworks:

If you only need basic model integration, RAG, Function Calling : use Spring AI 2.0 directly. It is the official Spring offering, most canonical ecosystem, and stable release cadence.

If you need complex workflow orchestration (e.g., Dify-style node-based flows) : Spring AI Alibaba's Graph capability remains usable; it is maintained but no longer aggressively iterated.

If you want "Agentic" applications where the model decides its own actions : watch AgentScope Java . This is where Alibaba's resource investment is concentrated, with extremely fast iteration.

The author notes that a previous recommendation of "Spring AI in Practice" assumed a context of basic AI capability integration , which remains valid — Spring AI is still the most orthodox entry point for Java developers accessing AI. However, for "Agent framework selection", the divergence between Spring AI Alibaba and AgentScope must be clearly explained, an area where the author's industry tracking was not timely enough.