Google ADK-Go (github.com)

Google has released ADK-Go, an open-source, code-first Agent Development Kit for building, evaluating and deploying AI agents natively in Go. Designed to feel idiomatic to Go developers, ADK-Go emphasizes concurrency and performance while applying software engineering best practices—defining agent logic, tools and orchestration directly in Go for better testability, versioning and CI/CD. The framework is modular and model-agnostic (though optimized for Gemini), supports pre-built and custom tools, and enables composition of scalable multi-agent systems. It’s deployment-agnostic and cloud-native friendly, with easy containerization and explicit support for environments like Google Cloud Run. For the AI/ML community, ADK-Go lowers the friction of productionizing agent-based systems in environments where Go already dominates—microservices, low-latency backends and infrastructure tooling. Key technical implications include first-class support for concurrent agent workflows, tighter integration with existing Go ecosystems, and cleaner operational patterns for orchestration and observability. ADK-Go is available via go get google.golang.org/adk and is released under the Apache 2.0 license (note an exception for internal/httprr in its LICENSE file), making it suitable for enterprises that need permissive licensing and cloud-native deployment workflows.