智能体开发工具包(ADK)

Vertex AI

Defi

多智能体

Agent2Agent协议: 合作伙伴生态系统，智能体间通信协议

MCP vs. A2A

AP2(智能体支付协议), OAuth, API密钥VC's, HTTP/SSL/DNSSEC

智能体Web:

Google Cloud Run支持GPU

Vertex AI Agent Engine

Firebase Studio

Gemini CLI

React优化

Jules - 异步开发智能体

Next.js

Google Cloud

Gemma 开放模型 - 可Fine Tune

Nano banana - Gemini 2.5 Flash Image ： 生图

Imagen 4 ： 高保真图像生成模型，生成高质量、高仿真图

Veo 3.1 是Google最先进的模型，生成高保真视频

Gemini 2.5 TTS - Notebook LLM同款

02 - Platforms

AI Studio

Vertex AI

Firebase AI logic

Server端：genai SDK接入Gemini API

Vertex AI: 企业级后台

客户端: firebase SDK 接入使用

Agent development Kit (ADK)

Firebase : 云端

Claude code + vertex AI - RMB 的力量

最顺手的Toolkit:

Cline + Gemini API + Vertex AI Claude

Not suggested:

• Cursor + AI Studio Gemini API --- 工具调用差
• Cursor + Vertex AI Claude

Gemini APP - Deep research

Gemini in Workspace, Firebase

DevTools MCP Server

MCP flow

Gemini CLI - open source CLI

Under the hood:

Googler demos

integrate

Stitch: Export to figma and code (front-end )

Google labs

deploy AI agent

log, monitor and trace

Agenda:

Build

Architect

Operate agents at scale

Agent - model

​ - Tools

Why build an agent?

How to build an agent?

Agent Development Kit.

fast, serverless

Deploy your agent to Google Cloud

MCP:

What value does MCP add?

Configure

LLMagent

SequentialAgent

A2A protocol

ADK callbacks:

Operate Agents

Gemma on Google Cloud Run

Cloud Run: Serverless Offering

Accelerate AI with Cloud Run

ADK write MCP server

Cloud shell Editor, Gemini CLI

Level1: AI使用

Level2: 开发APP, Agent,

Level3: 模型微调、训练

Learn GDG杭州 × Datawhale 谷歌开源学习计划

参与Datawhale的学习计划，学习AI