From Docker to LangGraph: Why the agent framework shakeout is the container wars all over again. Learn to bet on protocols, not frameworks, in this new era.

The New Stack is a publication covering trends and technologies in cloud-native development, DevOps, and software delivery. Developers can learn about containerization, Kubernetes, and cloud computing, as well as explore topics such as microservices architecture, serverless computing, and continuous integration/continuous delivery (CI/CD) pipelines.

The New Stack

The AI agent framework landscape mirrors the container orchestration wars of 2015, where hyperscalers (AWS, Google, Microsoft, OpenAI) are giving away open-source frameworks like Strands, ADK, and Agents SDK as on-ramps to their paid inference and deployment runtimes—the same GKE/EKS/AKS playbook applied to agents. Unlike containers, agent frameworks face a two-front squeeze: hyperscaler runtimes commoditize deployment from above, while increasingly capable LLMs reduce the need for complex orchestration from below. Real value is shifting to context engineering, evaluation/observability tooling, agent security, and interoperability protocols (MCP, A2A). The recommendation for platform teams: bet on protocols over frameworks, invest in framework-agnostic observability, and lean into hyperscaler SDKs if already cloud-committed. The 'Kubernetes of agents' may turn out to be the protocol layer, not any single framework.

The reason big tech is giving away AI agent frameworks