Announcing Genkit Middleware: Intercept, extend, and harden your agentic apps

2026-05-14 · Source: Google Developers Blog - AI · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Intermediate, short

Summary

Genkit, an open-source framework for building AI-powered and agentic applications, has introduced a new middleware system. This system provides composable hooks that intercept generation calls, including the tool execution loop, to inject custom behaviors for reliability, human approval, and observability. Middleware operates at three layers: Generate (per tool-loop iteration), Model (per model API call), and Tool (per tool execution). Genkit offers pre-built middleware for common use cases such as Retry for transient errors, Fallback to alternative models, Tool Approval for human-in-the-loop confirmation, Skills for injecting prompt content, and Filesystem for scoped model access. Developers can also create custom middleware to enforce specific application rules, like content filtering, and inspect their execution via the Genkit Developer UI.

Key takeaway

For AI Engineers building production-ready agentic applications, Genkit's new middleware system offers critical capabilities to harden your deployments. You should explore integrating pre-built middleware like Retry and Fallback for enhanced reliability, and consider developing custom middleware for application-specific logic such as content filtering or data governance. This approach centralizes control and improves the maintainability of complex AI workflows.

Key insights

Genkit middleware enhances agentic AI applications with composable hooks for reliability, control, and custom logic.

Principles

• Composable hooks enable custom behavior injection.
• Middleware operates at distinct layers of the tool loop.
• Order of middleware in a stack matters.

Method

Implement custom middleware by providing a name and a factory function that returns specific hooks (WrapModel, WrapGenerate, WrapTool) to intercept and modify AI generation calls.

In practice

• Use Retry middleware for model API call reliability.
• Implement Tool Approval for human-in-the-loop actions.
• Develop custom content filters for model outputs.

Topics

• Genkit Framework
• AI Agentic Applications
• Middleware System
• Tool Execution Loop
• Reliability Patterns

Code references

• genkit-ai/genkit

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• Presentation: Powering the Future: Building Your GenAI Infrastructure Stack — Intuit's GenOS platform and "fixed, flexible, free" framework accelerate enterprise-scale AI agent development through structured governance and continuous evaluation.
• Your Agent Loop Is Fine. Your Tools Are Why It Breaks. — Agent reliability hinges on well-defined tools, not complex orchestration loops.
• NirDiamant / agents-towards-production — This playbook offers practical, open-source tutorials for building and deploying production-grade Generative AI agents.
• The New Security Risks of the Agentic Development Lifecycle — Agentic development shifts security focus from code artifacts to the trustworthiness of the AI system creating them.
• 4 years of observing and living through the AI era — Generative and agentic AI tools offer unprecedented automation and creation capabilities, shifting the focus to human validation and decision-making.
• How Addepar Scales Investment Workflows with Databricks AI Agents — Unified data, governance, and agent orchestration are critical for secure, scalable GenAI in regulated financial services.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Google Developers Blog - AI.