skillpack.co
← All solutions

Google Agent Development Kit (ADK)

active

4.3M PyPI downloads/month for a framework under 10 months old. v1.27.2 (2026-03-17), multi-language (Python, TypeScript, Go, Java), model-agnostic (Gemini, Claude, Ollama, vLLM, LiteLLM). Native Cloud Run + Vertex AI Agent Engine deployment. ADK 2.0 Alpha adds graph-based workflows. Best for GCP/Vertex AI teams.

Score 92

Where it wins

4.4M PyPI downloads/month — fastest absolute star growth (18.5K in <12 months)

Multi-language: Python, TypeScript, Go, Java — widest language breadth in category

Model-agnostic despite Google origins: Gemini, Claude, Ollama, vLLM, LiteLLM

Pre-built Workflow agents (Sequential, Parallel, Loop) reduce boilerplate

Native Cloud Run + Vertex AI Agent Engine deployment — unique GCP advantage

Most complete DevOps story: built-in evaluation, testing, containerization, deployment pipelines

Named customers: Renault Group, Box, Revionics (Google blog, self-reported)

Where to be skeptical

GCP/Vertex lock-in concern — unfavorable tradeoff vs LangGraph for non-GCP teams

Bi-weekly release cadence may introduce instability

Named customers from Google-controlled publications only

Editorial verdict

Strong download velocity for its age — GCP-native deployment and multi-language commitment give it a longer runway than single-language frameworks. Best for teams already on GCP/Vertex AI. No independently-verified production case studies outside Google-controlled publications.

Videos

Reviews, tutorials, and comparisons from the community.

Introduction to Google ADK

Google·2025

Related

Teams of Agents / Multi-Agent Orchestration

#F06of 23

GCP-native teams wanting fastest path from prototype to deployed agent

Claude Code

98

Anthropic's official agentic coding CLI. v2.1.81 (Mar 20) shipped `--bare`, smarter worktree resume, and improved MCP OAuth while the repo crossed 82,204 stars and logged ~14 commits/week across 10+ maintainers. Terminal-native, tool-use-driven, with deep file system + shell access, #1 SWE-bench Pro standardized (45.89%), ~4% of GitHub public commits (SemiAnalysis), $2.5B annualized revenue. 8M+ npm weekly downloads. Opus 4.6 with 1M context.

LangGraph

95

#1 Python agent framework by production evidence — 40.2M PyPI downloads/month, Fortune 500 deployments (LinkedIn, Uber, Replit, Elastic, Klarna, Cloudflare, Coinbase), ~400 LangGraph Platform companies, LangSmith rated best-in-class observability. Stable v1.x API, model-agnostic, MCP support.

Pydantic AI

95

#3 Python agent framework by downloads — 15.6M PyPI/month. Built by the Pydantic team. Runtime type enforcement is a genuine differentiator no other framework offers. V1 shipped with Temporal integration for durable execution and Logfire observability. Emerging pattern: 'Pydantic AI for agent logic, LangGraph for orchestration' (ZenML).

AutoGen (Microsoft)

95

⚠️ MAINTENANCE MODE — Microsoft officially confirmed bug fixes and security patches only, no new features (VentureBeat 2026-02-19). 55.9K stars but only 1.57M PyPI/month — DL/star ratio of 28, the most inflated among active frameworks. Being replaced by Microsoft Agent Framework (AutoGen + Semantic Kernel merge, GA targeted ~Q2 2026). Teams on AutoGen should plan migration.

Public evidence

Raw GitHub source

GitHub README peek

Constrained peek so you can sanity-check the source material without leaving the site.

Agent Development Kit (ADK) 2.0

<h2 align="center"> <img src="https://raw.githubusercontent.com/google/adk-python/main/assets/agent-development-kit.png" width="256"/> </h2> <h3 align="center"> An open-source, code-first Python framework for building, evaluating, and deploying sophisticated AI agents with flexibility and control. </h3> <h3 align="center"> Important Links: <a href="https://google.github.io/adk-docs/">Docs</a>, <a href="https://github.com/google/adk-samples">Samples</a> & <a href="https://github.com/google/adk-web">ADK Web</a>. </h3>

⚠️ BREAKING CHANGES FROM 1.x

This release includes breaking changes to the agent API, event model, and session schema. Sessions generated by ADK 2.0 are readable by ADK 1.28+ (extra fields will be ignored), but are incompatible with older 1.x versions.

🔥 What's New in 2.0

  • Workflow Runtime: A graph-based execution engine for composing deterministic execution flows for agentic apps, with support for 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.

🚀 Installation

pip install google-adk

Requirements: Python 3.11+.

To install optional integrations, you can use the following command:

pip install "google-adk[extensions]"

The release cadence is roughly bi-weekly.

Quick Start

Agent
from google.adk import Agent

root_agent = Agent(
    name="greeting_agent",
    model="gemini-2.5-flash",
    instruction="You are a helpful assistant. Greet the user warmly.",
)
Workflow
from google.adk import Agent, Workflow

generate_fruit_agent = Agent(
    name="generate_fruit_agent",
    instruction="Return the name of a random fruit. Return only the name.",
)

generate_benefit_agent = Agent(
    name="generate_benefit_agent",
    instruction="Tell me a health benefit about the specified fruit.",
)

root_agent = Workflow(
    name="root_agent",
    edges=[("START", generate_fruit_agent, generate_benefit_agent)],
)
Run Locally
# Interactive CLI
adk run path/to/my_agent

# Web UI (supports multi-agent directories or pointing directly to a single agent folder)
adk web path/to/agents_dir

📚 Documentation

  • Getting Started: https://google.github.io/adk-docs/
  • Samples: See contributing/workflow_samples/ and contributing/task_samples/ for workflow and task API examples.

🤝 Contributing

See CONTRIBUTING.md for details.

📄 License

This project is licensed under the Apache 2.0 License — see the LICENSE file for details.

View on GitHub →