Google Genkit Middleware Gives Developers Runtime Control Over AI Apps
Google has launched a middleware system for Genkit, its open-source framework for building AI-powered and agentic applications. The update brings a programmable interception layer that wraps around model calls, tool execution, and generation loops. As a result, developers now gain more direct control over reliability, safety, and orchestration in production AI systems.
The Google Genkit middleware release lets developers inject custom behaviors into workflows without touching core application logic. Teams can now add retries, model fallbacks, and logging at the middleware level. Genkit already supports TypeScript, Go, and Dart. Python support is also coming soon.
Every generate() call in Genkit runs through a tool loop. In this loop, the model generates output, executes tools, processes results, and continues until completion. According to Google, the new middleware hooks can intercept that cycle at three levels: generation, model calls, and tool execution. This gives teams fine-grained control at each stage.
Google also shipped several prebuilt middleware components alongside the update. These include retry handling with exponential backoff and automatic fallback to alternative models when APIs fail. Additionally, teams get approval gates for sensitive tool calls, filesystem access controls, and a “skills” system that injects instructions dynamically from local files.
Furthermore, middleware components can stack together. Retries, filters, approvals, and logging all operate in a defined execution order. The system also connects to the Genkit Developer UI, where developers can inspect middleware behavior, trace execution flows, and debug runtime interactions in real time.
The release reflects a broader shift in AI tooling. Rather than relying purely on prompts or model tuning, frameworks are increasingly adding programmable runtime layers to govern how models behave in production. Google Genkit middleware fits squarely into this trend.
The announcement also sparked debate on X about how Genkit relates to Google’s Agent Development Kit (ADK). Michael Doyle, a software engineer at Google, stepped in to clarify the distinction. He explained that Genkit targets developers adding agentic features to existing apps, while ADK is built for complex, standalone multi-agent systems running on dedicated infrastructure like GCP’s Agent Platform.
In short, Google is positioning Genkit as an application-layer framework. It integrates AI features into existing products, whereas ADK handles heavier orchestration workloads.
Developers can start using the Google Genkit middleware system now through the latest Genkit release. They can also publish custom middleware packages for reuse across projects.
