Google Genkit Agents API Preview: Build Full-Stack Conversational AI Agents with TypeScript/Go

Google has released a preview of the Agents API for the Genkit framework. It supports TypeScript and Go, allowing developers to handle everything from server-side agent definition to streaming conversations via the chat() interface, state management, tool loops, human approval, and HTTP serving in a unified way. Integration with Firebase and Firestore makes it suitable for production use.

📑Table of Contents

1. Overview and Background of Genkit Agents API
2. Core Features of the Agents API (Definition, State, Streaming, Interrupts)
3. Full-Stack Integration and Client Connectivity
4. Differences from ADK and Selection Criteria
5. Genkit Agents vs Traditional generate() Approach
6. Frequently Asked Questions (FAQ)
7. Summary

---

Overview and Background of Genkit Agents API

Genkit is Google’s framework for building AI applications. Previously, the simple generate() call was the norm. With the Agents API, developers can now systematically build complex agent logic on the server side. According to the official Google Developers Blog, you define agents using DefineAgent with system prompts, tools, and state. The feature has gained attention on X as a potential new standard for agentic development.

---

Core Features of the Agents API (Definition, State, Streaming, Interrupts)

The heart of the Agents API lies in the DefineAgent function. You specify a system prompt, available tools, and a session store to create an agent. Conversations run through RunText or the streaming-enabled chat() interface.

A key capability is support for human-in-the-loop interrupts during tool execution. Session stores include FirestoreSessionStore, in-memory, file-based, or custom options. Long-running tasks can be detached and polled later.

Here is a Go example for defining an agent:

agent := genkit.DefineAgent("my-agent", systemPrompt, tools, FirestoreSessionStore{})

TypeScript offers equivalent definitions. The Developer UI Agent Runner makes testing conversations and interrupts straightforward.

---

Full-Stack Integration and Client Connectivity

A major strength is full-stack support. AllAgentRoutes lets you build an HTTP server, while remoteAgent on the frontend uses the same chat() interface.

This enables natural combinations of JavaScript/TypeScript frontends with Go or TypeScript backends. Delegation to sub-agents and artifact handling are also possible. Firebase integration provides production-grade session persistence.

---

Differences from ADK and Selection Criteria

Google’s ADK (Agent Development Kit) and the Agents API serve different purposes. ADK focuses on complex multi-agent orchestration and managed runtimes. In contrast, Genkit Agents API excels at embedding agents directly within applications.

Selection guidelines:

• Single-app embedded agent control → Genkit Agents API
• Complex multi-agent workflow management → ADK

For Claude Code and Cursor users, the ability to extend existing Genkit experience with agent features is particularly appealing.

---

Genkit Agents vs Traditional generate() Approach

Aspect	Traditional generate()	Genkit Agents API
Agent Definition	Simple prompt invocation	DefineAgent with tools, state, and sessions
Streaming	Basic support	Advanced streaming and interrupts via chat()
Human Approval	Manual implementation required	Standard support via InterruptibleTool
State Management	Developer implements manually	Official stores (Firestore etc.) for simplicity
Full-Stack	Separate HTTP implementation needed	Unified via AllAgentRoutes + remoteAgent
Production Integration	Custom handling	Native Firebase/Firestore support

Source: Google Developers Blog (as of July 2026)

---

Frequently Asked Questions (FAQ)

Q: Is the Genkit Agents API ready for production use?

It is in preview, but Firestore integration and HTTP serving provide a solid foundation for production. Watch for possible breaking changes in minor releases.

Q: Can I use stores other than Firestore?

Yes. In-memory, file-based, and custom stores are available. In-memory is convenient for development, while Firestore is recommended for production.

Q: How does the human approval mechanism work?

InterruptibleTool allows confirmation before tool execution. Official documentation includes examples such as shell command execution.

Q: Is migration from existing Genkit projects straightforward?

You can introduce it gradually by adding DefineAgent. Coexistence with generate() is supported.

---

Related articles:

• Sony Ends aibo Domestic Sales in Japan | Inventory Clearance, Ongoing Support
• Understanding Debt in the AI Coding Era: Developer Survival Strategies
• Why a Cheap Keychain Outperforms Expensive AI Smart Glasses for Everyday Reminders

Summary

The preview release of the Genkit Agents API significantly simplifies building full-stack conversational AI agents. By understanding the TypeScript/Go support, state management, interrupts, and HTTP integration based on official sources, Claude Code and Cursor users can apply these capabilities in real projects.

Start by reviewing the official Agents overview documentation and the Google Developers Blog announcement, then experiment with small agents. Leveraging Firebase integration enables efficient development of production-grade agent applications.

Related new article:

• How NotebookLM Streamlines Requirements Definition — Preparing for Specification-Driven Development (SDD) – This published update adds current operational context for Google Genkit Agents API Preview: Build Full-Stack Conversational AI Agents with TypeScript/Go.
• Optimized Hostname Routing Tokyo App to Ohio DB — Direct Hostname Fix – This published update adds current operational context for Google Genkit Agents API Preview: Build Full-Stack Conversational AI Agents with TypeScript/Go.
• How to Exit Failing Projects: Practical Steps to Avoid Sunk Cost Fallacy in IT and Development – This published update adds current operational context for Google Genkit Agents API Preview: Build Full-Stack Conversational AI Agents with TypeScript/Go.
• Kanto’s Largest Electronics Experience Space — June 30 2026 Grand Opening Comprehensive Guide – This published update adds current operational context for Google Genkit Agents API Preview: Build Full-Stack Conversational AI Agents with TypeScript/Go.
• Personal Task Management Design: Human Judgment, Agent Updates, Script Calculations – This published update adds current operational context for Google Genkit Agents API Preview: Build Full-Stack Conversational AI Agents with TypeScript/Go.
• Ray-Ban Meta Smart Glasses: Hands-Free Kids Moments Capture for Parents 2026 – This published update adds current operational context for Google Genkit Agents API Preview: Build Full-Stack Conversational AI Agents with TypeScript/Go.
• High-Density Fluid Enables Pumped Hydro Storage Without Mountains – This published update adds current operational context for Google Genkit Agents API Preview: Build Full-Stack Conversational AI Agents with TypeScript/Go.
• GPT-5.6 Sol vs Terra vs Luna: Pricing, Strengths, and How to Start – This published update adds current operational context for Google Genkit Agents API Preview: Build Full-Stack Conversational AI Agents with TypeScript/Go.
• GPT-5.6 Sol Token Usage and API Cost Estimation Guide (2026) – This published update adds current operational context for Google Genkit Agents API Preview: Build Full-Stack Conversational AI Agents with TypeScript/Go.