top of page

LBSocial

OpenClaw + Google Workspace CLI:构建管理 Gmail、文档与课堂的 AI 智能体

已更新:2天前

别再只是和 AI 聊天了——开始构建一个真正能干活的智能体吧。在本教程中,我们将强大的 OpenClaw 框架与全新的 Google Workspace CLI 集成,创建一个完全自主的办公助手。我们不仅限于简单的文本生成,还将展示如何为您的 AI 智能体安装“双手”,让它能够跨 Gmail、云端硬盘、文档和 Google Classroom 管理您的职业生活。


⚠️ 注意:Google Workspace CLI 是一个开源项目,并非 Google 官方支持的产品。

Telegram sends tasks to Gemini AI on Google Cloud via OpenClaw, which accesses Google Workspace services. Cute crab AI illustration.
The OpenClaw workflow: Sending tasks via Telegram to your AI agent, which directly accesses Google Workspace using the installed CLI skills.


Watch the full video tutorial here:

OpenClaw + Google Workspace CLI: Build an AI Agent for Gmail, Docs, and Classroom


夯实 OpenClaw 基础


在执行 Google Workspace 集成之前,您需要先运行起云端环境和智能体。如果您尚未设置基础架构,强烈建议从我们的基础教程开始:



掌握了这些核心概念后,您就可以将 OpenClaw 连接到 Google Workspace 了!


Split screen showing a Google Classroom page with a Python intro and a chat in a messaging app discussing an announcement about a GitHub document.
In this tutorial, we build an agent that manages Google Docs and Google Classroom directly via Telegram commands.

理解架构:双账号策略


为了确保自动化流程的安全,我们将工作流拆分为两个账号:


  1. 开发者账号 (Developer Account):用于设置 Google Cloud 项目并生成 "Client Secret"(应用程序的身份证)。

  2. 用户账号 (User Account):用于授权实际访问 Workspace 数据。


这种分离确保了您的智能体仅能访问特定的演示账号,从而保护您的主数据免受意外删除。


Flowchart comparing Google Cloud project setup and user authorization, featuring a cartoon crab, icons for APIs, and green authorization steps.
Conceptual Chart: Google Cloud Project Setup vs. User Authorization.


第一阶段:Google Cloud 设置与 API 配置


在使用任何命令之前,我们需要先利用开发者账号打好云端基础。


  1. 创建专用项目:在 Google Cloud Console 中,专为 Workspace CLI 创建一个新项目。将其与 OpenClaw VM 项目分开是最佳实践。

  2. 启用 API:前往“API 与服务”库,启用智能体需要的特定应用。在本教程中,请启用 Google Drive, Gmail, Calendar, Google Docs, Sheets, Slides, Tasks 和 Google Classroom 的 API。


Google Cloud console showing Google Classroom API details as enabled. Notifications list several enabled services like Sheets and Docs.
Manually enabling the required Google Workspace APIs in your new Google Cloud Project.

  1. 配置 OAuth 同意屏幕:在身份验证平台中,用户类型选择 External(外部)。填写应用名称和开发者联系邮箱。

  2. 添加测试用户:在“受众”或“测试用户”部分,输入您用户账号(演示账号)的邮箱地址。

  3. 生成 Client Secret:前往“凭据”,点击“创建凭据”,选择 OAuth 客户端 ID。应用类型选择 桌面应用 (Desktop App)。创建后,下载 JSON 文件。这就是您的 Client Secret。


第二阶段:本地电脑授权


由于运行 OpenClaw 的远程虚拟机 (VM) 没有浏览器,我们必须先在本地电脑上生成 OAuth 授权令牌。


Local computer authorizes apps via OAuth, creating a .json token sent to a virtual machine. Blue and white design with arrows indicating flow.
Passing the generated OAuth token from your Local Computer to the Virtual Machine.


首先,创建 CLI 存放凭据的隐藏配置文件夹,并将从 Google Cloud 下载的 Client Secret 文件复制进去:


# 1. Create the hidden config directory on your local machine
mkdir "%USERPROFILE%\.config\gws"

# 2. Copy your downloaded Google Cloud secret to this new folder
copy "your_downloaded_client_secret.json" "%USERPROFILE%\.config\gws\client_secret.json"

接下来,启动登录流程。此命令将安全地打开您的网页浏览器,允许您使用演示用户账号登录并授予特定权限:


# 3. Request authorization for specific Workspace apps via your browser
gws auth login --scopes "https://www.googleapis.com/auth/drive,https://www.googleapis.com/auth/gmail.modify,https://www.googleapis.com/auth/calendar,https://www.googleapis.com/auth/documents,https://www.googleapis.com/auth/spreadsheets,https://www.googleapis.com/auth/presentations,https://www.googleapis.com/auth/tasks,https://www.googleapis.com/auth/classroom.courses,https://www.googleapis.com/auth/classroom.courseworkmaterials,https://www.googleapis.com/auth/classroom.announcements"

Popup showing "OpenClaw Workspace Agent" requesting Google Account access, with options for Drive, Slides, Docs, and more on the right.
Authorizing the specific scopes via the browser on your local machine.

在浏览器点击“允许”后,检查状态并将最终授权的令牌导出到新文件:


# 4. Verify successful login
gws auth status

# 5. Export the authorized token to a new JSON file
gws auth export --unmasked > credentials.json


第三阶段:远程虚拟机配置


现在,切换到安装了 OpenClaw 的远程 Linux 虚拟机。您需要创建一个配置文件夹,将刚才生成的 JSON 令牌内容粘贴进去,并设置文件权限以确保安全。


# 1. Create the config directory on the remote VM
mkdir -p ~/.config/gws

# 2. Open a text editor to paste the contents of your exported credentials.json
nano ~/.config/gws/credentials.json

SSH browser interface shows GNU nano editing credentials.json. Gray background with JSON fields. Toolbar at the bottom includes commands.
 Securely pasting the authorized credentials into the remote VM using the nano editor.

# 3. Tell the CLI exactly where to find your credentials file
export GOOGLE_WORKSPACE_CLI_CREDENTIALS_FILE=~/.config/gws/credentials.json

# 4. Restrict file permissions so only the owner can read the token
chmod 600 ~/.config/gws/credentials.json

如果您尚未在虚拟机上安装 Google Workspace CLI,可以使用 Rust/Cargo 从源码构建:


# Install required dependencies and Rust
sudo apt update
sudo apt install -y curl build-essential pkg-config libssl-dev git
curl https://sh.rustup.rs -sSf | sh -s -- -y
source "$HOME/.cargo/env"

# Build and install the Google Workspace CLI
cargo install --git https://github.com/googleworkspace/cli --locked

# Verify the VM is successfully authenticated with Google Workspace
gws auth status

Text lines in a terminal window showing URLs related to Google Classroom API permissions, with SSH-in-browser header and toolbar.
Verifying that the remote VM has successfully authenticated using the uploaded credentials.

第四阶段:安装 OpenClaw Skills


有了令牌和工具,最后我们需要教 OpenClaw 如何使用 Google Workspace CLI。我们将克隆仓库并将预定义的“技能”复制到 OpenClaw 的活跃技能文件夹中。


# 1. Navigate to the home directory and clone the CLI repository
cd ~
git clone https://github.com/googleworkspace/cli.git googleworkspace-cli

# 2. Ensure the OpenClaw skills directory exists
mkdir -p ~/.openclaw/skills

# 3. Copy all the Workspace skills into OpenClaw's system
cp -r ~/googleworkspace-cli/skills/gws-* ~/.openclaw/skills/

# 4. Verify the skills are recognized by OpenClaw
openclaw skills list

Console display with text in columns, showing "ready" status and Google Workspace features like admin reports and calendar. Dark background.
Confirming OpenClaw has successfully recognized and loaded the Google Workspace skills.

结语:测试您的 AI 智能体


执行完这些命令后,您的 OpenClaw 智能体现在拥有了管理您数字工作区的直接授权访问权限。启动 OpenClaw (openclaw),打开 Telegram,给您的智能体下达指令吧!


Google Classroom and messaging app split screen. Left: Python Programming Spring 2026 class stream with announcements. Right: Chat about class setup.
Testing the integration by commanding the agent to draft a new Google Doc from Telegram.

您可以要求智能体总结未读邮件、安排日历事件,或者将新生成的文档直接附加到 Google Classroom 的作业中。


留言


bottom of page