Agent Development Kit (ADK)

Overview

Agent Development Kit (ADK) is an open-source, code-first Python framework for building, evaluating, and deploying sophisticated AI agents. It applies software development principles to AI agent creation and is designed to simplify building, deploying, and orchestrating agent workflows, from simple tasks to complex systems. ADK is optimized for Gemini but is model-agnostic, deployment-agnostic, and compatible with other frameworks.

Key Features

• Workflow Runtime: graph-based execution engine for composing deterministic agentic app flows with routing, fan-out/fan-in, loops, retry, state management, dynamic nodes, human-in-the-loop, and nested workflows.
• Task API: structured agent-to-agent delegation with multi-turn task mode, single-turn controlled output, mixed delegation patterns, human-in-the-loop, and task agents as workflow nodes.
• Modular Multi-Agent Systems: compose multiple specialized agents into flexible hierarchies for scalable applications.
• Rich Tool Ecosystem: use pre-built tools, custom functions, OpenAPI specs, MCP tools, or existing tools to give agents diverse capabilities.
• Code-First Development: define agent logic, tools, and orchestration directly in Python for flexibility, testability, and versioning.
• Deploy Anywhere: containerize and deploy agents on Cloud Run or scale with Vertex AI Agent Engine.

Use Cases

• Building, evaluating, and deploying sophisticated AI agents with flexibility and control.
• Creating modular multi-agent systems by composing specialized agents into flexible hierarchies.
• Orchestrating agent workflows from simple tasks to complex systems with routing, loops, retry, state, and human-in-the-loop.
• Running local development and evaluation through interactive CLI, Web UI, and bundled evaluation samples.

Getting Started

• Install the latest stable release with pip install google-adk; Python 3.10+ is required.
• For transitive dependency protection, install with companion constraints files for Python 3.10 to 3.14.
• Define an Agent with name, model, and instruction, and define a Workflow with edges to orchestrate agents.
• Run locally with adk run path/to/my_agent or use the Web UI with adk web path/to/agents_dir.
• Optional integrations can be installed with pip install "google-adk[extensions]".

Deployment & Requirements

• Python 3.10+ is required.
• Agents can be deployed locally in a Docker container with adk deploy docker --with_ui.
• Agents can be deployed in Google Cloud with Cloud Run using adk deploy cloud_run --with_ui.
• Environment variables may be set in the adk command or in a .env file, for example GOOGLE_GENAI_USE_ENTERPRISE=1.
• Vertex AI Agent Engine is mentioned as a scaling option.

Before You Adopt

• License: Apache-2.0. Review its terms before using, modifying, or distributing the project.
• The development version may contain experimental changes or bugs, so it is intended mainly for testing upcoming changes or accessing critical fixes before official release.