ADK Go

Google has announced the release of Agent Development Kit (ADK) for Go 1.0, marking a major step toward making AI agents production-ready for cloud and enterprise environments.

The update brings advanced features like observability, plugins, and human-in-the-loop controls, signaling a shift from experimental AI scripts to scalable, real-world agent systems.

Built for Production-Grade AI Agents

With ADK Go 1.0, Google is focusing on the realities of deploying AI agents in production—where reliability, debugging, and control are critical.

The framework enables developers to build:

👉 The goal: make AI agents behave like real software systems, not just prototypes.

OpenTelemetry Integration for Deep Observability

One of the biggest additions is native OpenTelemetry (OTel) support, which gives developers visibility into how agents operate internally.

With this, teams can:

This addresses a major challenge in AI systems: understanding why an agent made a decision.

Plugin System Enables “Self-Healing” Agents

ADK Go 1.0 introduces a new plugin system that allows developers to extend agent behavior without modifying core logic.

Notably, Google highlights a “Retry and Reflect” plugin, which:

👉 This creates self-healing agents that can fix their own mistakes automatically.

Human-in-the-Loop for Safety and Control

To address safety concerns, the framework includes Human-in-the-Loop (HITL) capabilities.

Developers can require approval for sensitive actions such as:

The agent pauses execution until a human confirms the action, adding a critical layer of governance.

YAML-Based Configuration for Faster Iteration

Another key feature is support for YAML-based agent configuration, allowing teams to define agents without writing full Go code.

This enables:

👉 Developers can modify agent behavior without rebuilding applications.

Multi-Language Agent Collaboration (A2A Protocol)

ADK Go 1.0 also improves support for the Agent2Agent (A2A) protocol, enabling agents written in different languages to communicate seamlessly.

This means:

Why Go Matters for AI Agents

Go is widely used for:

By bringing ADK to Go, Google is targeting developers building high-throughput, production-grade AI systems that need performance and concurrency.