Java Ecosystem Roundup: A2A SDK, Spring AI M5, and Gradle 9.5
The InfoQ Java News Roundup for the week of April 27, 2026 packed in several releases worth tracking. The most notable was Agent2Agent Java SDK 1.0.0-Beta1, the first beta release of the Java client for Google's A2A inter-agent communication protocol. The SDK supports multiple transport protocols and gives Java developers a typed, idiomatic entry point into multi-agent systems — a critical gap now that A2A has been donated to the Linux Foundation and is gaining adoption across enterprise frameworks.
Spring AI 2.0.0-M5 shipped alongside the roundup with improved structured output converters and new builder methods for message construction. Structured output has been a pain point in production Spring AI integrations, and M5's converters make it easier to bind LLM responses directly to POJOs without manual parsing. The fifth milestone brings Spring AI closer to a general availability release later this year.
Gradle 9.5.0 introduced enhanced task diagnostics, giving developers a clearer picture of which tasks are running, why they're incremental, and where time is being spent. The release also ships a build-time dependency tree shaker alongside Quarkus 3.35.0, which removes unused dependencies at compile time to reduce container image sizes — a meaningful improvement for teams running microservices on serverless runtimes.
Additional releases in the roundup: Eclipse GlassFish 9.0.0 M2 added JDK 25 compatibility and early implementations of Jakarta Security 5.0-M2 and Jakarta Faces 5.0-M2; JReleaser 1.24.0 introduced Zernio announcer support; and LangChain4j 1.14.0 expanded its AI service feature set for building typed LLM pipelines in Java.
Read more — InfoQ
Java Annotated Monthly May 2026: AI Harnesses and the Agentic Development Shift
JetBrains' monthly Java digest for May 2026 leads with a theme that will resonate with any team now using AI tooling day-to-day: how to design effective harnesses for AI coding agents so they produce reliable output rather than plausible-looking mistakes. The curation argues that the highest-leverage skill for Java developers in 2026 is not learning yet another framework but instead designing the feedback loops — tests, constraints, and verification steps — that keep AI agents on track.
The digest highlights practical patterns for embedding AI agents into Spring-based backends, including stateful continuation architectures that allow agents to pause, checkpoint, and resume long-running workflows without losing context. Test-driven development is reframed as the natural "harness" for agents: a passing test suite is the ground truth an agent optimizes toward, making TDD more relevant, not less, in the agentic era.
The May edition also surfaces the Java conference circuit: JAX, Devoxx UK, GeeCon, and JNation all have Java and AI tracks this spring, offering community-driven perspectives on where the ecosystem is actually heading versus where vendor roadmaps say it is.
Read more — The IntelliJ IDEA Blog
Links & Sources
You May Also Enjoy
