ADK Agent Deployment Samples
This repository contains sample agents and scripts for deploying them using the Agent Development Kit (ADK).
📺 You can watch the accompanying YouTube Video.
接入你的智能体
從發布者安裝。不是 PlugMCP 端點。
遠端 URL 還是本機指令? 兩種安裝方式的差異
來自發布者
This repository contains sample agents and scripts for deploying them using the Agent Development Kit (ADK).
📺 You can watch the accompanying YouTube Video.
This project shows how to deploy Agent Development Kit (ADK) agents that use Model Context Protocol (MCP) servers onto Google Cloud using Cloud Run. I
For a comprehensive crash course on Google ADK covering Agent2Agent (A2A) Protocol and Model Context Protocol (MCP), check out the 1.5-hour tutorial that goes into much more detail about setting up ADK agents.
🎓 Watch the FREE crash course: YouTube Video
I have compiled a complete toolkit for building production-ready ADK agents, including:
.cursor/rules for use with AI code editors🚀 Get it now: ADK Builder Pack
If you find this tutorial series and codebase helpful in your AI agent development journey, consider buying me a coffee! Your support helps me create more educational content on AI and agent development.
Google Cloud SDK (gcloud): Ensure you have gcloud installed and authenticated. You can find instructions here.
Project Root Dockerfile.template: A Dockerfile.template file must exist in the root of this project. This template is used to generate a specific Dockerfile for the agent. It should look something like this:
# Use the official ADK base image
FROM us-docker.pkg.dev/agent-development-kit/adk-images/adk-agent-base:latest
# Copy the agent-specific code and dependencies
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt
# The __AGENT_DIR__ placeholder will be replaced by the script
COPY agents/__AGENT_DIR__ /app/agents/__AGENT_DIR__
# Set the agent directory environment variable for the ADK
ENV AGENT_DIR=__AGENT_DIR__
# Run the agent
CMD ["adk", "run"]
First, you need to set the necessary environment variables for your Google Cloud project.
export GOOGLE_CLOUD_PROJECT="your-gcp-project-id"
export GOOGLE_CLOUD_LOCATION="your-gcp-region" # e.g., us-central1
Replace "your-gcp-project-id" and "your-gcp-region" with your actual Google Cloud project ID and desired region.
To ensure a clean build, we'll create a temporary directory that will contain only the files needed for our agent.
mkdir temp_staging
cd temp_staging
Copy the agent code and the general requirements into the staging directory.
# Create the directory structure gcloud expects
mkdir agents
# Copy the agent code
cp -r ../multitool_agent ./agents/
# Copy the project's requirements
cp ../requirements.txt .
# Ensure Python treats the 'agents' directory as a package
touch agents/__init__.py
The gcloud run deploy --source command looks for a Dockerfile in the directory you point it to. We need to create one from our Dockerfile.template.
You will need to manually create the Dockerfile from Dockerfile.template. Open ../Dockerfile.template, replace the __AGENT_DIR__ placeholder with multitool_agent, and save the new file as Dockerfile inside the temp_staging directory.
Your temp_staging directory should now look like this:
temp_staging/
├── Dockerfile
├── agents/
│ ├── __init__.py
│ └── multitool_agent/
│ ├── __init__.py
│ └── agent.py
└── requirements.txt
Now, from the root of the project (not inside temp_staging), run the deployment command. We'll name our service multitool-agent-service.
# Make sure you are in the project's root directory
cd ..
# Deploy!
gcloud run deploy multitool-agent-service \
--source temp_staging \
--region "$GOOGLE_CLOUD_LOCATION" \
--project "$GOOGLE_CLOUD_PROJECT" \
--memory=1Gi \
--allow-unauthenticated
Google Cloud Build will now use the contents of the temp_staging directory to build a container image and deploy it to Cloud Run.
Once the deployment is successful, you can remove the temporary staging directory.
rm -rf temp_staging
When you run a deployment, gcloud may prompt you to allow unauthenticated invocations. To run these scripts in a non-interactive way (e.g., in a CI/CD pipeline), you can set the default behavior for your project:
gcloud run services update-iam-policy-binding [SERVICE_NAME] \
--region=[REGION] \
--member="allUsers" \
--role="roles/run.invoker" \
--project=[PROJECT_ID]
Replace [SERVICE_NAME], [REGION], and [PROJECT_ID] with your specific service name, region, and project ID. This command grants public access to the service, which is a common requirement for public-facing web services or APIs.
This project shows how to deploy Agent Development Kit (ADK) agents that use Model Context Protocol (MCP) servers onto Google Cloud using Cloud Run. I
工具
尚無工具清單。遠端伺服器會在首次健康檢查後取得工具清單。
常見問題
在本頁複製適用於 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 設定。它使用發布者的安裝方式,而不是 PlugMCP 代理伺服器。
這筆記錄未列出必要的環境變數。
Adk Deploy Agents Samples 透過 stdio 在本機執行。
上次檢查時間:2026-10-28。 This project shows how to deploy Agent Development Kit (ADK) agents that use Model Context Protocol (MCP) servers onto Google Cloud using Cloud Run. I