---
title: "Why we built ADK 2.0"
description: "ADK 2.0 addresses the structural challenge of moving AI agents from prototypes into production, where infinite loops, hallucinated bypasses, and failures without clean exceptions make model-led orchestration unreliable. Because language models are slow, costly, and variable for routing, scheduling, and error handling, the release separates execution routing from language processing and reserves LLMs for ambiguous or cognitive steps. Building on ADK v1’s support for Python, Java, Go, TypeScript, and Kotlin, it adds a structured workflow runtime and task-collaboration model, with Workflows available in Python since March and newly launched for Go. Deterministic graphs combine tool calls, Human-in-the-Loop steps, and specialized agents with explicit branching and error handling. Dynamic workflows extend this model with native Python and asyncio control flows for retries, branching, and reusable sub-workflows."
---

# Why we built ADK 2.0

[Google](https://yomu.fyi/company/google) · Swapnil Agarwal, Alan Blount, Frank Guan · Jul 1, 2026

**Type:** Explainer

## Summary

ADK 2.0 addresses the structural challenge of moving AI agents from prototypes into production, where infinite loops, hallucinated bypasses, and failures without clean exceptions make model-led orchestration unreliable. Because language models are slow, costly, and variable for routing, scheduling, and error handling, the release separates execution routing from language processing and reserves LLMs for ambiguous or cognitive steps. Building on ADK v1’s support for Python, Java, Go, TypeScript, and Kotlin, it adds a structured workflow runtime and task-collaboration model, with Workflows available in Python since March and newly launched for Go. Deterministic graphs combine tool calls, Human-in-the-Loop steps, and specialized agents with explicit branching and error handling. Dynamic workflows extend this model with native Python and asyncio control flows for retries, branching, and reusable sub-workflows.

## Context

Moving AI agents into real-world enterprise production introduces infinite loops, hallucinated bypasses of business logic, failures without clean exceptions, and unnecessary cost and latency when LLMs perform deterministic orchestration. Developers need both the flexibility of agents for ambiguous tasks and the predictability of traditional workflow code.

## Approach / What changed

ADK 2.0 introduces a structured workflow runtime and task-collaboration model. Its Workflows separate execution routing from language processing, combine deterministic graph nodes with LLM agents and Human-in-the-Loop steps, and support conditional routing, dynamic Python and asyncio control flows, modular sub-workflows, and specialized collaboration modes.

## Takeaways

- ADK 2.0 Workflows represent business processes as deterministic directed graphs, allowing tool calls and other fixed steps to execute without an LLM deciding the next route.
- The refund example confines LLM use to policy analysis and email drafting while programmatic nodes fetch purchase history, issue refunds, and update the support ticket.
- The shown comparison lists 5,152 tokens per run for a vanilla LLM agent and 2,265 for an ADK 2.0 Workflow, with approximately 50% savings.

**Tags:** [Agent Development Kit](https://yomu.fyi/topic/agent-development-kit), [AI Agents](https://yomu.fyi/topic/ai-agents), [Architecture](https://yomu.fyi/topic/architecture), [Python](https://yomu.fyi/topic/python)

- Source: [Google](https://developers.googleblog.com/why-we-built-adk-20)
- Source URL: https://developers.googleblog.com/why-we-built-adk-20
- Ingested by Yomu: 2026-09-09T14:29:02.794Z

[Read original post](https://developers.googleblog.com/why-we-built-adk-20)