Google released ADK for Java 1.0.0 on March 30 and ADK for Go 1.0 on March 31, bringing its Agent Development Kit to four language SDKs: Python, TypeScript, Java, and Go. No other major agent framework supports more than two languages. For enterprise Java and Go teams that have been forced to maintain separate Python stacks for agent development, the releases eliminate a layer of tooling friction.
The Go 1.0 release adds native OpenTelemetry integration for distributed tracing, a plugin system with self-healing retry logic, and human-in-the-loop confirmation workflows. The Java release ships with A2A protocol support, event compaction for managing context window sizes, and session persistence via Vertex AI and Firestore. Both include grounding tools like GoogleMapsTool and UrlContextTool for connecting agents to external data.
The Two-Camp Split
A Morph comparison published on April 5 cataloging eight agent frameworks identifies a structural divide in the landscape. On one side: provider-native SDKs from Anthropic (Claude Agent SDK), OpenAI (Agents SDK), and Google (ADK). These offer deep integration with their respective model families, including features like Anthropic’s 200+ MCP server ecosystem and OpenAI’s built-in guardrail system. On the other: independent frameworks like LangGraph, CrewAI (45,900+ GitHub stars, 5.2 million monthly downloads according to Morph), Smolagents, Pydantic AI, and AutoGen that work across model providers but lack native protocol support.
The trade-off is concrete. Claude Agent SDK has the deepest MCP integration in the ecosystem but locks developers to Claude models. OpenAI’s Agents SDK offers the cleanest multi-agent handoff pattern but has no built-in state persistence. Google’s ADK has the widest language support and native A2A, but carries a heavier Google Cloud dependency for production deployments. Independent frameworks like LangGraph offer crash recovery and time-travel debugging that no provider SDK matches, but require adapter layers for protocol support.
Protocol Consolidation
The protocol layer underneath these frameworks is converging. IBM launched the Agent Communication Protocol (ACP) in March 2025 to power its BeeAI platform. Google announced A2A the following month. In August 2025, the Linux Foundation announced ACP would merge into A2A, according to Virtua Cloud. StackOne confirmed the merger completed in early 2026, noting that A2A now has 150+ supporting organizations and added gRPC support. MCP, meanwhile, has crossed 200 server implementations after Anthropic donated it to the Agentic AI Foundation under the Linux Foundation in December 2025.
The practical result for builders: agent-to-agent communication is consolidating around A2A, while tool access is consolidating around MCP. CrewAI is currently the only independent framework with native support for both protocols. Google ADK has native A2A but relies on adapters for MCP. Anthropic’s Claude Agent SDK has the deepest MCP but no A2A support at all.
What It Means for Builders
The framework choice increasingly comes down to a bet on where integration depth matters most. Teams committed to a single model provider get the best developer experience from that provider’s native SDK. Teams running multi-model deployments or planning to swap providers need the portability of independent frameworks, accepting thinner protocol integration as the cost. The ADK Java and Go releases make Google the first provider to seriously compete for the enterprise teams that were never going to rewrite their backends in Python.