DEV Community: xbill

MCP Development with Gemini CLI, and Amazon AWS EKS

xbill — Tue, 07 Apr 2026 19:57:35 +0000

Leveraging Gemini CLI and the underlying Gemini LLM to build Model Context Protocol (MCP) AI applications with Python with a local development environment hosted on AWS EKS.

Aren’t There a Billion Python MCP Demos?

Yes there are.

Python has traditionally been the main coding language for ML and AI tools. The goal of this article is to provide a minimal viable basic working MCP stdio server that can be run locally without any unneeded extra code or extensions.

What Is Python?

Python is an interpreted language that allows for rapid development and testing and has deep libraries for working with ML and AI:

Welcome to Python.org

Python Version Management

One of the downsides of the wide deployment of Python has been managing the language versions across platforms and maintaining a supported version.

The pyenv tool enables deploying consistent versions of Python:

GitHub - pyenv/pyenv: Simple Python version management

As of writing — the mainstream python version is 3.13. To validate your current Python:

admin@ip-172-31-70-211:~/gemini-cli-aws/mcp-stdio-python-aws$ python --version
Python 3.13.12

Amazon EKS

Amazon Elastic Kubernetes Service (EKS) is a fully managed service from Amazon Web Services (AWS) that makes it easy to run Kubernetes on AWS without needing to install, operate, or maintain your own Kubernetes control plane. It automates cluster management, security, and scaling, supporting applications on both Amazon EC2 and AWS Fargate.

More information is available here:

What is Amazon EKS?

Gemini CLI

If not pre-installed you can download the Gemini CLI to interact with the source files and provide real-time assistance:

npm install -g @google/gemini-cli

Testing the Gemini CLI Environment

Once you have all the tools and the correct Node.js version in place- you can test the startup of Gemini CLI. You will need to authenticate with a Key or your Google Account:

gemini

admin@ip-172-31-70-211:~/gemini-cli-aws/mcp-stdio-python-aws$ gemini

▝▜▄ Gemini CLI v0.33.1
    ▝▜▄
   ▗▟▀ Logged in with Google /auth
  ▝▀ Gemini Code Assist Standard /upgrade

? for shortcuts 
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 shift+tab to accept edits 3 GEMINI.md files | 1 MCP server
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 > Type your message or @path/to/file
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 ~/.../mcp-stdio-python-aws (main*) no sandbox (see /docs) /model Auto (Gemini 3) | 239.8 MB

Node Version Management

Gemini CLI needs a consistent, up to date version of Node. The nvm command can be used to get a standard Node environment:

GitHub - nvm-sh/nvm: Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions

Python MCP Documentation

The official GitHub Repo provides samples and documentation for getting started:

GitHub - modelcontextprotocol/python-sdk: The official Python SDK for Model Context Protocol servers and clients

The most common MCP Python deployment path uses the FASTMCP library:

Welcome to FastMCP - FastMCP

Where do I start?

The strategy for starting MCP development is a incremental step by step approach.

First, the basic development environment is setup with the required system variables, and a working Gemini CLI configuration.

Then, a minimal Hello World Style Python MCP Server is built with stdio transport. This server is validated with Gemini CLI in the local environment.

This setup validates the connection from Gemini CLI to the local process via MCP. The MCP client (Gemini CLI) and the Python MCP server both run in the same local environment.

Next- the basic MCP server is extended with Gemini CLI to add several new tools in standard Python code.

Setup the Basic Environment

At this point you should have a working Python interpreter and a working Gemini CLI installation. The next step is to clone the GitHub samples repository with support scripts:

cd ~
git clone https://github.com/xbill9/gemini-cli-aws

Then run init.sh from the cloned directory.

The script will attempt to determine your shell environment and set the correct variables:

cd gemini-cli-aws
source init.sh

If your session times out or you need to re-authenticate- you can run the set_env.sh script to reset your environment variables:

cd gemini-cli-aws
source set_env.sh

Variables like PROJECT_ID need to be setup for use in the various build scripts- so the set_env script can be used to reset the environment if you time-out.

Hello World with HTTP Transport

One of the key features that the standard MCP libraries provide is abstracting various transport methods.

The high level MCP tool implementation is the same no matter what low level transport channel/method that the MCP Client uses to connect to a MCP Server.

The simplest transport that the SDK supports is the stdio (stdio/stdout) transport — which connects a locally running process. Both the MCP client and MCP Server must be running in the same environment.

The HTTP transport allows the MCP Client and Server to be in the same environment or distributed over the Internet.

The connection over HTTP will look similar to this:

mcp.run(
        transport="http",
        host="0.0.0.0",
        port=port,
    )

Running the Python Code

First- switch the directory with the Python MCP sample code:

cd ~/gemini-cli-aws/mcp-https-python-eks

Run the release version on the local system:

cd ~/gemini-cli-aws/mcp-https-python-eks
make release

You can validate the final result by checking the messages:

✦ The release target has been successfully executed, and the server is running locally on port 8080.

  The health check confirms that it is responding:

   1 {"status":"healthy","service":"mcp-server"}

  The process is running in the background with PID 29456.

Deploy to EKS

Now that the server has been tested locally- start the remote deployment to EKS.

Use the deploy target:

 > make deploy
✦ I will now execute the make deploy target to rebuild the Docker image, push it to ECR, and update the Kubernetes
  deployment on Amazon EKS.

Then check the status:

✦ The EKS deployment is fully operational, with the pod READY and the LoadBalancer service accessible via the
  external endpoint.

  Status Summary:
   - Deployment: mcp-server-image (1/1 pods available)
   - External Endpoint: http://a8acf5c3ad24348c89537b575c29b32f-1627118381.us-east-1.elb.amazonaws.com
   - Port: 80

And the endpoint:

✦ The EKS LoadBalancer public endpoint is:
  a8acf5c3ad24348c89537b575c29b32f-1627118381.us-east-1.elb.amazonaws.com

Gemini CLI settings.json

The default Gemini CLI settings.json has an entry for the Python source:

   1 {
   2 "mcpServers": {
   3 "mcp-eks-python-aws": {
   4 "httpUrl": "http://a8acf5c3ad24348c89537b575c29b32f-1627118381.us-east-1.elb.amazonaws.com/mcp"
   5 }
   6 }
   7 }

Validation with Gemini CLI

Leave the HTTP server running and open a new terminal.

Gemini CLI is restarted and the MCP connection over HTTP to the Python Code is validated. The full Gemini CLI Session will start:

admin@ip-172-31-70-211:~/gemini-cli-aws/mcp-https-python-aws$ gemini

▝▜▄ Gemini CLI v0.33.1
    ▝▜▄
   ▗▟▀ Logged in with Google /auth
  ▝▀ Gemini Code Assist Standard /upgrade

 > /mcp list
Configured MCP servers:

🟢 mcp-eks-python-aws - Ready (1 tool)
  Tools:
  - mcp_mcp-eks-python-aws_greet

And you can then connect to the MCP Server over HTTP:

                                                                    │
 > greet eks
✦ Calling the greet tool with the argument "eks".

╭────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ ✓ greet (mcp-eks-python-aws MCP Server) {"param":"eks"} │
│ │
│ eks │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
✦ The greet tool returned: eks

Project Review

The full deployment was reviewed with Gemini CLI:

✦ I have completed my review of the Amazon EKS project setup. The repository is well-structured and follows
  industry best practices for containerized Python applications on Kubernetes.

  Key Strengths
   - Infrastructure as Code: The use of eksctl with eks-cluster.yaml provides a clear and reproducible way to
     manage the EKS cluster.
   - Automation: The Makefile is comprehensive, covering the entire development lifecycle from local testing and
     linting to Docker builds and EKS deployment.
   - Observability: Structured JSON logging is implemented correctly, which is essential for effective log
     analysis in CloudWatch.
   - Reliability: The inclusion of Kubernetes liveness and readiness probes linked to a dedicated /health endpoint
     ensures the service remains stable in production.
   - Security: The eks-cluster.yaml is pre-configured for IAM Roles for Service Accounts (IRSA) via OIDC, allowing
     for fine-grained permissions if needed in the future.

Summary

The strategy for using Python for MCP development with Gemini CLI and AWS was validated with a incremental step by step approach.

A minimal HTTP transport MCP Server was started from Python source code and validated with Gemini CLI running as a MCP client in the same local environment.

The HTTP server was then tested remotely from the AWS EKS deployment.

This approach can be extended to more complex deployments using other MCP transports and Cloud based options.

Building a Multimodal Cross Cloud Live Agent with ADK, Azure AKS, and Gemini CLI

xbill — Tue, 07 Apr 2026 13:16:14 +0000

Leveraging the Google Agent Development Kit (ADK) and the underlying Gemini LLM to build cross cloud apps with the Python programming language deployed to the Azure app service.

Aren’t There a Billion Python MCP Demos?

Yes there are.

What Is Python?

Python is an interpreted language that allows for rapid development and testing and has deep libraries for working with ML and AI:

Welcome to Python.org

Python Version Management

One of the downsides of the wide deployment of Python has been managing the language versions across platforms and maintaining a supported version.

The pyenv tool enables deploying consistent versions of Python:

GitHub - pyenv/pyenv: Simple Python version management

As of writing — the mainstream python version is 3.13. To validate your current Python:

admin@ip-172-31-70-211:~/gemini-cli-azure$ python --version
Python 3.13.12

Azure Kubernates Service

Azure Kubernetes Service (AKS) is a fully managed, serverless Kubernetes service on Microsoft Azure that simplifies deploying, scaling, and managing containerized applications. It handles critical tasks like health monitoring, maintenance, and automated upgrades, reducing operational complexity. AKS is used for microservices, DevOps, and cloud-native app development.

More details are available here:

https://azure.microsoft.com/en-us/products/kubernetes-service

Isn’t that Overkill? A whole Cluster Just for some Agents?!

An entire cluster is a large deployment for just a basic ADK server. The goal was to validate that ADK servers can be deployed — and that opens the door for more complex deployments that can take advantage of the full services in the cluster.

Why would I want Gemini CLI with Azure? Isn’t that a Google Thing?

Yes- Gemini CLI leverages the Google Cloud console and Gemini models but it is also open source and platform agnostic. Many applications are already cross-cloud so this enables familiar tools to be run natively on Microsoft Azure.

Gemini CLI

If not pre-installed you can download the Gemini CLI to interact with the source files and provide real-time assistance:

npm install -g @google/gemini-cli

Testing the Gemini CLI Environment

Once you have all the tools and the correct Node.js version in place- you can test the startup of Gemini CLI. You will need to authenticate with a Key or your Google Account:

gemini

admin@ip-172-31-70-211:~/gemini-cli-azure$ gemini

▝▜▄ Gemini CLI v0.33.1
    ▝▜▄
   ▗▟▀ Logged in with Google /auth
  ▝▀ Gemini Code Assist Standard /upgrade

? for shortcuts 
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 shift+tab to accept edits 3 GEMINI.md files | 1 MCP server
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 > Type your message or @path/to/file
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 ~/.../gemini-cli-azure (main*) no sandbox (see /docs) /model Auto (Gemini 3) | 239.8 MB

Node Version Management

Gemini CLI needs a consistent, up to date version of Node. The nvm command can be used to get a standard Node environment:

GitHub - nvm-sh/nvm: Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions

Docker Version Management

The Azure CLI tools need current version of Docker. If your environment does not provide a recent docker tool- the Docker Version Manager can be used to downlaod the latest supported Docker:

Install

Azure CLI

The Azure CLI provides a command line tool to directly access Azure services from your current environment. Full details on the CLI are available here:

Azure Command-Line Interface (CLI) documentation

Agent Development Kit

The Google Agent Development Kit (ADK) is an open-source, Python-based framework designed to streamline the creation, deployment, and orchestration of sophisticated, multi-agent AI systems. It treats agent development like software engineering, offering modularity, state management, and built-in tools (like Google Search) to build autonomous agents.

The ADK can be installed from here:

Agent Development Kit (ADK)

This seems like a lot of Configuration!

Getting the key tools in place is the first step to working across Cloud environments. For a deeper dive- a project with a similar setup can be found here:

MCP Development with Python, and the Azure App Service

Where do I start?

The strategy for starting multimodal real time cross cloud agent development is a incremental step by step approach.

The agents in the demo are based on the original code lab:

Way Back Home - Building an ADK Bi-Directional Streaming Agent | Google Codelabs

First, the basic development environment is setup with the required system variables, and a working Gemini CLI configuration.

Then, a minimal ADK Agent is built with the visual builder. Next — the entire solution is deployed to Google Cloud Run.

Setup the Basic Environment

At this point you should have a working Python environment and a working Gemini CLI installation. All of the relevant code examples and documentation is available in GitHub. This repo has a wide variety of samples- but this lab will focus on the ‘level_3-lightsail’ setup.

The next step is to clone the GitHub repository to your local environment:

cd ~
git clone https://github.com/xbill9/gemini-cli-azure
cd level_3-aks

Then run init.sh from the cloned directory.

The script will attempt to determine your shell environment and set the correct variables:

source init.sh

If your session times out or you need to re-authenticate- you can run the set_env.sh script to reset your environment variables:

source set_env.sh

Variables like PROJECT_ID need to be setup for use in the various build scripts- so the set_env script can be used to reset the environment if you time-out.

Verify The ADK Installation

To verify the setup, run the ADK CLI locally with Agent1:

xbill@penguin:~/gemini-cli-azure/level_3-aks/backend/app$ adk run biometric_agent
Log setup complete: /tmp/agents_log/agent.20260403_091821.log
To access latest log: tail -F /tmp/agents_log/agent.latest.log
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/cli.py:204: UserWarning: [EXPERIMENTAL] InMemoryCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  credential_service = InMemoryCredentialService()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/auth/credential_service/in_memory_credential_service.py:33: UserWarning: [EXPERIMENTAL] BaseCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  super(). __init__ ()
Running agent biometric_agent, type exit to exit.



  
  
  Deploying to Azure Kubernates Service


The first step is to refresh the Azure credentials in the current build environment:



xbill@penguin:~/gemini-cli-azure/level_3-appservice$ az login






Run the deploy version on the local system:




xbill@penguin:~/gemini-cli-azure/level_3-aks$ make deploy
./deploy-aks.sh
Ensuring Resource Group exists... 0.0s 0.0s






You can validate the final result by checking the messages:




xbill@penguin:~/gemini-cli-azure/level_3-aks$ make status
Checking AKS status for biometric-scout-aks...
Name State K8sVersion
------------------- --------- ------------
biometric-scout-aks Succeeded 1.33
Kubernetes Services:
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
biometric-scout-service ClusterIP 10.0.252.23 <none> 80/TCP 68m
Kubernetes Pods:
NAME READY STATUS RESTARTS AGE
biometric-scout-app-5bb549f585-dpwrx 1/1 Running 0 68m
xbill@penguin:~/gemini-cli-azure/level_3-aks$ 






The service will be visible in the Azure console:




  
  
  Running the Web Interface


Start a connection to the deployed app:




  https://biometric-scout-app.orangebeach-e5cba679.canadaeast.azurecontainerapps.io






Then connect tothe app :



Then use the Live model to process audio and video:



Finally — complete the sequence:




  
  
  Summary


The Agent Development Kit was used to enable a multi-modal agent using the Gemini Live Model. This Agent was tested locally with the CLI and then deployed to Azure Kubernates services. This approach validates that cross cloud tools can be used — even with more complex agents.

MCP Development with Gemini CLI, and Amazon AWS EKS

xbill — Tue, 07 Apr 2026 13:14:40 +0000

Leveraging Gemini CLI and the underlying Gemini LLM to build Model Context Protocol (MCP) AI applications with Python with a local development environment hosted on AWS EKS.

Aren’t There a Billion Python MCP Demos?

Yes there are.

What Is Python?

Python is an interpreted language that allows for rapid development and testing and has deep libraries for working with ML and AI:

Welcome to Python.org

Python Version Management

One of the downsides of the wide deployment of Python has been managing the language versions across platforms and maintaining a supported version.

The pyenv tool enables deploying consistent versions of Python:

GitHub - pyenv/pyenv: Simple Python version management

As of writing — the mainstream python version is 3.13. To validate your current Python:

admin@ip-172-31-70-211:~/gemini-cli-aws/mcp-stdio-python-aws$ python --version
Python 3.13.12

Amazon EKS

More information is available here:

What is Amazon EKS?

Gemini CLI

If not pre-installed you can download the Gemini CLI to interact with the source files and provide real-time assistance:

npm install -g @google/gemini-cli

Testing the Gemini CLI Environment

Once you have all the tools and the correct Node.js version in place- you can test the startup of Gemini CLI. You will need to authenticate with a Key or your Google Account:

gemini

admin@ip-172-31-70-211:~/gemini-cli-aws/mcp-stdio-python-aws$ gemini

▝▜▄ Gemini CLI v0.33.1
    ▝▜▄
   ▗▟▀ Logged in with Google /auth
  ▝▀ Gemini Code Assist Standard /upgrade

? for shortcuts 
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 shift+tab to accept edits 3 GEMINI.md files | 1 MCP server
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 > Type your message or @path/to/file
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 ~/.../mcp-stdio-python-aws (main*) no sandbox (see /docs) /model Auto (Gemini 3) | 239.8 MB

Node Version Management

Gemini CLI needs a consistent, up to date version of Node. The nvm command can be used to get a standard Node environment:

GitHub - nvm-sh/nvm: Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions

Python MCP Documentation

The official GitHub Repo provides samples and documentation for getting started:

GitHub - modelcontextprotocol/python-sdk: The official Python SDK for Model Context Protocol servers and clients

The most common MCP Python deployment path uses the FASTMCP library:

Welcome to FastMCP - FastMCP

Where do I start?

The strategy for starting MCP development is a incremental step by step approach.

First, the basic development environment is setup with the required system variables, and a working Gemini CLI configuration.

Then, a minimal Hello World Style Python MCP Server is built with stdio transport. This server is validated with Gemini CLI in the local environment.

This setup validates the connection from Gemini CLI to the local process via MCP. The MCP client (Gemini CLI) and the Python MCP server both run in the same local environment.

Next- the basic MCP server is extended with Gemini CLI to add several new tools in standard Python code.

Setup the Basic Environment

At this point you should have a working Python interpreter and a working Gemini CLI installation. The next step is to clone the GitHub samples repository with support scripts:

cd ~
git clone https://github.com/xbill9/gemini-cli-aws

Then run init.sh from the cloned directory.

The script will attempt to determine your shell environment and set the correct variables:

cd gemini-cli-aws
source init.sh

If your session times out or you need to re-authenticate- you can run the set_env.sh script to reset your environment variables:

cd gemini-cli-aws
source set_env.sh

Variables like PROJECT_ID need to be setup for use in the various build scripts- so the set_env script can be used to reset the environment if you time-out.

Hello World with HTTP Transport

One of the key features that the standard MCP libraries provide is abstracting various transport methods.

The high level MCP tool implementation is the same no matter what low level transport channel/method that the MCP Client uses to connect to a MCP Server.

The HTTP transport allows the MCP Client and Server to be in the same environment or distributed over the Internet.

The connection over HTTP will look similar to this:

mcp.run(
        transport="http",
        host="0.0.0.0",
        port=port,
    )

Running the Python Code

First- switch the directory with the Python MCP sample code:

cd ~/gemini-cli-aws/mcp-https-python-eks

Run the release version on the local system:

cd ~/gemini-cli-aws/mcp-https-python-eks
make release

You can validate the final result by checking the messages:

✦ The release target has been successfully executed, and the server is running locally on port 8080.

  The health check confirms that it is responding:

   1 {"status":"healthy","service":"mcp-server"}

  The process is running in the background with PID 29456.

Deploy to EKS

Now that the server has been tested locally- start the remote deployment to EKS.

Use the deploy target:

 > make deploy
✦ I will now execute the make deploy target to rebuild the Docker image, push it to ECR, and update the Kubernetes
  deployment on Amazon EKS.

Then check the status:

✦ The EKS deployment is fully operational, with the pod READY and the LoadBalancer service accessible via the
  external endpoint.

  Status Summary:
   - Deployment: mcp-server-image (1/1 pods available)
   - External Endpoint: http://a8acf5c3ad24348c89537b575c29b32f-1627118381.us-east-1.elb.amazonaws.com
   - Port: 80

And the endpoint:

✦ The EKS LoadBalancer public endpoint is:
  a8acf5c3ad24348c89537b575c29b32f-1627118381.us-east-1.elb.amazonaws.com

Gemini CLI settings.json

The default Gemini CLI settings.json has an entry for the Python source:

   1 {
   2 "mcpServers": {
   3 "mcp-eks-python-aws": {
   4 "httpUrl": "http://a8acf5c3ad24348c89537b575c29b32f-1627118381.us-east-1.elb.amazonaws.com/mcp"
   5 }
   6 }
   7 }

Validation with Gemini CLI

Leave the HTTP server running and open a new terminal.

Gemini CLI is restarted and the MCP connection over HTTP to the Python Code is validated. The full Gemini CLI Session will start:

admin@ip-172-31-70-211:~/gemini-cli-aws/mcp-https-python-aws$ gemini

▝▜▄ Gemini CLI v0.33.1
    ▝▜▄
   ▗▟▀ Logged in with Google /auth
  ▝▀ Gemini Code Assist Standard /upgrade

 > /mcp list
Configured MCP servers:

🟢 mcp-eks-python-aws - Ready (1 tool)
  Tools:
  - mcp_mcp-eks-python-aws_greet

And you can then connect to the MCP Server over HTTP:

                                                                    │
 > greet eks
✦ Calling the greet tool with the argument "eks".

╭────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ ✓ greet (mcp-eks-python-aws MCP Server) {"param":"eks"} │
│ │
│ eks │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
✦ The greet tool returned: eks

Project Review

The full deployment was reviewed with Gemini CLI:

✦ I have completed my review of the Amazon EKS project setup. The repository is well-structured and follows
  industry best practices for containerized Python applications on Kubernetes.

  Key Strengths
   - Infrastructure as Code: The use of eksctl with eks-cluster.yaml provides a clear and reproducible way to
     manage the EKS cluster.
   - Automation: The Makefile is comprehensive, covering the entire development lifecycle from local testing and
     linting to Docker builds and EKS deployment.
   - Observability: Structured JSON logging is implemented correctly, which is essential for effective log
     analysis in CloudWatch.
   - Reliability: The inclusion of Kubernetes liveness and readiness probes linked to a dedicated /health endpoint
     ensures the service remains stable in production.
   - Security: The eks-cluster.yaml is pre-configured for IAM Roles for Service Accounts (IRSA) via OIDC, allowing
     for fine-grained permissions if needed in the future.

Summary

The strategy for using Python for MCP development with Gemini CLI and AWS was validated with a incremental step by step approach.

A minimal HTTP transport MCP Server was started from Python source code and validated with Gemini CLI running as a MCP client in the same local environment.

The HTTP server was then tested remotely from the AWS EKS deployment.

This approach can be extended to more complex deployments using other MCP transports and Cloud based options.

MCP Development with Python, Gemini CLI, and Amazon AWS ECS Express

xbill — Mon, 06 Apr 2026 17:18:50 +0000

Leveraging Gemini CLI and the underlying Gemini LLM to build Model Context Protocol (MCP) AI applications with Python with a local development environment hosted on AWS ECS Express.

Aren’t There a Billion Python MCP Demos?

Yes there are.

What Is Python?

Python is an interpreted language that allows for rapid development and testing and has deep libraries for working with ML and AI:

Welcome to Python.org

Python Version Management

One of the downsides of the wide deployment of Python has been managing the language versions across platforms and maintaining a supported version.

The pyenv tool enables deploying consistent versions of Python:

GitHub - pyenv/pyenv: Simple Python version management

As of writing — the mainstream python version is 3.13. To validate your current Python:

admin@ip-172-31-70-211:~/gemini-cli-aws/mcp-stdio-python-aws$ python --version
Python 3.13.12

Amazon ECS Express Configuration

Amazon ECS Express Mode (announced Nov 2025) is a simplified deployment feature for Amazon Elastic Container Service (ECS) designed to rapidly launch containerized applications, APIs, and web services on AWS Fargate. It automates infrastructure setup — including load balancing, networking, scaling, and HTTPS endpoints — allowing developers to deploy from container image to production in a single step.

More details are available here:

Amazon ECS Express Mode

Gemini CLI

If not pre-installed you can download the Gemini CLI to interact with the source files and provide real-time assistance:

npm install -g @google/gemini-cli

Testing the Gemini CLI Environment

Once you have all the tools and the correct Node.js version in place- you can test the startup of Gemini CLI. You will need to authenticate with a Key or your Google Account:

gemini

admin@ip-172-31-70-211:~/gemini-cli-aws/mcp-stdio-python-aws$ gemini

▝▜▄ Gemini CLI v0.33.1
    ▝▜▄
   ▗▟▀ Logged in with Google /auth
  ▝▀ Gemini Code Assist Standard /upgrade

? for shortcuts 
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 shift+tab to accept edits 3 GEMINI.md files | 1 MCP server
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 > Type your message or @path/to/file
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 ~/.../mcp-stdio-python-aws (main*) no sandbox (see /docs) /model Auto (Gemini 3) | 239.8 MB

Node Version Management

Gemini CLI needs a consistent, up to date version of Node. The nvm command can be used to get a standard Node environment:

GitHub - nvm-sh/nvm: Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions

Python MCP Documentation

The official GitHub Repo provides samples and documentation for getting started:

GitHub - modelcontextprotocol/python-sdk: The official Python SDK for Model Context Protocol servers and clients

The most common MCP Python deployment path uses the FASTMCP library:

Welcome to FastMCP - FastMCP

Where do I start?

The strategy for starting MCP development is a incremental step by step approach.

First, the basic development environment is setup with the required system variables, and a working Gemini CLI configuration.

Then, a minimal Hello World Style Python MCP Server is built with stdio transport. This server is validated with Gemini CLI in the local environment.

This setup validates the connection from Gemini CLI to the local process via MCP. The MCP client (Gemini CLI) and the Python MCP server both run in the same local environment.

Next- the basic MCP server is extended with Gemini CLI to add several new tools in standard Python code.

Setup the Basic Environment

At this point you should have a working Python interpreter and a working Gemini CLI installation. The next step is to clone the GitHub samples repository with support scripts:

cd ~
git clone https://github.com/xbill9/gemini-cli-aws

Then run init.sh from the cloned directory.

The script will attempt to determine your shell environment and set the correct variables:

cd gemini-cli-aws
source init.sh

If your session times out or you need to re-authenticate- you can run the set_env.sh script to reset your environment variables:

cd gemini-cli-aws
source set_env.sh

Variables like PROJECT_ID need to be setup for use in the various build scripts- so the set_env script can be used to reset the environment if you time-out.

Hello World with HTTP Transport

One of the key features that the standard MCP libraries provide is abstracting various transport methods.

The high level MCP tool implementation is the same no matter what low level transport channel/method that the MCP Client uses to connect to a MCP Server.

The HTTP transport allows the MCP Client and Server to be in the same environment or distributed over the Internet.

The connection over HTTP will look similar to this:

mcp.run(
        transport="http",
        host="0.0.0.0",
        port=port,
    )

Running the Python Code

First- switch the directory with the Python MCP sample code. Then, run the release version on the local system:

cd ~/gemini-cli-aws/mcp-https-python-ecs-express
make release

If everything is running correctly — you will see the FASTMCP banner:


                 ╭──────────────────────────────────────────────────────────────────────────────╮                 
                 │ │                 
                 │ │                 
                 │ ▄▀▀ ▄▀█ █▀▀ ▀█▀ █▀▄▀█ █▀▀ █▀█ │                 
                 │ █▀ █▀█ ▄▄█ █ █ ▀ █ █▄▄ █▀▀ │                 
                 │ │                 
                 │ │                 
                 │ FastMCP 3.2.0 │                 
                 │ https://gofastmcp.com │                 
                 │ │                 
                 │ 🖥 Server: hello-world-server, 3.2.0 │                 
                 │ 🚀 Deploy free: https://horizon.prefect.io │                 
                 │ │                 
                 ╰──────────────────────────────────────────────────────────────────────────────╯                 

[04/04/26 19:10:50] INFO Starting MCP server 'hello-world-server' with transport 'http' on transport.py:299
                             http://0.0.0.0:8080/mcp                                                              
INFO: Started server process [27502]
INFO: Waiting for application startup.
{"message": "StreamableHTTP session manager started"}
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8080 (Press CTRL+C to quit)

At this point — hit Cntrl-C to exit the local HTTP server.

Deploy to ECS Express

Now that the server has been tested locally- start the remote deployment to ECS Express.

Use the deploy target:

 > make deploy
✦ I will execute the full deployment cycle, which includes building the Docker image, pushing it to ECR, and
  deploying to ECS Express Mode.

Then check the status:

 > make status
✦ I will check the status of the ECS Express Mode service.

╭────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ ✓ Shell make status [in /home/xbill/gemini-cli-aws/mcp-ecsexpress-python-aws] │
│ │
│ Checking AWS ECS service status for mcp-express-python-aws... │
│ Service: mcp-express-python-aws │
│ Status: ACTIVE │
│ Endpoint: mc-8d69e7dfa87344be98fdb3f7a8fbbbba.ecs.us-east-1.on.aws │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
✦ The ECS Express Mode service mcp-express-python-aws is ACTIVE and available at:
  mc-8d69e7dfa87344be98fdb3f7a8fbbbba.ecs.us-east-1.on.aws

And the endpoint:

  Based on the make status output earlier, the actual endpoint is:
  mc-8d69e7dfa87344be98fdb3f7a8fbbbba.ecs.us-east-1.on.aws

Gemini CLI settings.json

The default Gemini CLI settings.json has an entry for the Python source:

{
  "mcpServers": {
    "aws-ecs-express-python": {
      "httpUrl": "https://mc-8d69e7dfa87344be98fdb3f7a8fbbbba.ecs.us-east-1.on.aws/mcp"
    }
  }
}

Validation with Gemini CLI

Leave the HTTP server running and open a new terminal.

Gemini CLI is restarted and the MCP connection over HTTP to the Python Code is validated. The full Gemini CLI Session will start:


xbill@penguin:~/gemini-cli-aws/mcp-ecsexpress-python-aws$ gemini

 ▝▜▄ Gemini CLI v0.36.0
   ▝▜▄
  ▗▟▀ Signed in with Google: xbill@glitnir.com /auth
 ▝▀ Plan: Gemini Code Assist Standard /upgrade

 > /mcp list
Configured MCP servers:

🟢 aws-ecs-express-python - Ready (1 tool)
  Tools:
  - mcp_aws-ecs-express-python_greet

And you can then connect to the MCP Server over HTTP:

 > greet ECS Express!
✦ I will call the greet tool with "ECS Express!".

╭────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ ✓ greet (aws-ecs-express-python MCP Server) {"param":"ECS Express!"} │
│ │
│ ECS Express! │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
✦ The greet tool returned: ECS Express!

Validate Project Setup with Gemini CLI

Gemini CLI was used to summarize the project:

 > summarize the project deployment to Amazon ECS Express
✦ The deployment of the mcp-express-python-aws project to Amazon ECS Express Mode is a streamlined process that
  automates the container lifecycle from build to public endpoint.

  Deployment Workflow (make deploy)
   1. Containerization: Builds a Python 3.13-slim Docker image and pushes it to an Amazon ECR repository
      (mcp-server-repo).
   2. Identity & Access: Automatically provisions necessary IAM roles (ecsTaskExecutionRole and
      ecsInfrastructureRoleForExpressServices) if they do not already exist.
   3. Orchestration: Deploys the container to ECS Express Mode, which simplifies AWS Fargate deployments by
      automatically managing the Application Load Balancer, Security Groups, and Networking.

  Current Infrastructure State
   * Service Name: mcp-express-python-aws
   * Status: ACTIVE
   * Compute: AWS Fargate (Task size: 1 vCPU / 2GB RAM)
   * Scaling: Configured for 1 to 20 tasks, triggered at 60% average CPU usage.
   * Public Access: mc-8d69e7dfa87344be98fdb3f7a8fbbbba.ecs.us-east-1.on.aws (Port 8080)

  The service is currently running and reachable via its assigned ECS Express Gateway endpoint.

Summary

The strategy for using Python for MCP development with Gemini CLI and AWS was validated with a incremental step by step approach.

A minimal HTTP transport MCP Server was started from Python source code and validated with Gemini CLI running as a MCP client in the same local environment.

The HTTP server was then tested remotely from the AWS ECS Express instance.

This approach can be extended to more complex deployments using other MCP transports and Cloud based options.

MCP Development with Python, Gemini CLI, and Amazon AWS ECS Express

xbill — Mon, 06 Apr 2026 13:09:29 +0000

Leveraging Gemini CLI and the underlying Gemini LLM to build Model Context Protocol (MCP) AI applications with Python with a local development environment hosted on AWS ECS Express.

Aren’t There a Billion Python MCP Demos?

Yes there are.

What Is Python?

Python is an interpreted language that allows for rapid development and testing and has deep libraries for working with ML and AI:

Welcome to Python.org

Python Version Management

One of the downsides of the wide deployment of Python has been managing the language versions across platforms and maintaining a supported version.

The pyenv tool enables deploying consistent versions of Python:

GitHub - pyenv/pyenv: Simple Python version management

As of writing — the mainstream python version is 3.13. To validate your current Python:

admin@ip-172-31-70-211:~/gemini-cli-aws/mcp-stdio-python-aws$ python --version
Python 3.13.12

Amazon ECS Express Configuration

More details are available here:

Amazon ECS Express Mode

Gemini CLI

If not pre-installed you can download the Gemini CLI to interact with the source files and provide real-time assistance:

npm install -g @google/gemini-cli

Testing the Gemini CLI Environment

Once you have all the tools and the correct Node.js version in place- you can test the startup of Gemini CLI. You will need to authenticate with a Key or your Google Account:

gemini

admin@ip-172-31-70-211:~/gemini-cli-aws/mcp-stdio-python-aws$ gemini

▝▜▄ Gemini CLI v0.33.1
    ▝▜▄
   ▗▟▀ Logged in with Google /auth
  ▝▀ Gemini Code Assist Standard /upgrade

? for shortcuts 
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 shift+tab to accept edits 3 GEMINI.md files | 1 MCP server
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 > Type your message or @path/to/file
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 ~/.../mcp-stdio-python-aws (main*) no sandbox (see /docs) /model Auto (Gemini 3) | 239.8 MB

Node Version Management

Gemini CLI needs a consistent, up to date version of Node. The nvm command can be used to get a standard Node environment:

GitHub - nvm-sh/nvm: Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions

Python MCP Documentation

The official GitHub Repo provides samples and documentation for getting started:

GitHub - modelcontextprotocol/python-sdk: The official Python SDK for Model Context Protocol servers and clients

The most common MCP Python deployment path uses the FASTMCP library:

Welcome to FastMCP - FastMCP

Where do I start?

The strategy for starting MCP development is a incremental step by step approach.

First, the basic development environment is setup with the required system variables, and a working Gemini CLI configuration.

Then, a minimal Hello World Style Python MCP Server is built with stdio transport. This server is validated with Gemini CLI in the local environment.

This setup validates the connection from Gemini CLI to the local process via MCP. The MCP client (Gemini CLI) and the Python MCP server both run in the same local environment.

Next- the basic MCP server is extended with Gemini CLI to add several new tools in standard Python code.

Setup the Basic Environment

At this point you should have a working Python interpreter and a working Gemini CLI installation. The next step is to clone the GitHub samples repository with support scripts:

cd ~
git clone https://github.com/xbill9/gemini-cli-aws

Then run init.sh from the cloned directory.

The script will attempt to determine your shell environment and set the correct variables:

cd gemini-cli-aws
source init.sh

If your session times out or you need to re-authenticate- you can run the set_env.sh script to reset your environment variables:

cd gemini-cli-aws
source set_env.sh

Variables like PROJECT_ID need to be setup for use in the various build scripts- so the set_env script can be used to reset the environment if you time-out.

Hello World with HTTP Transport

One of the key features that the standard MCP libraries provide is abstracting various transport methods.

The high level MCP tool implementation is the same no matter what low level transport channel/method that the MCP Client uses to connect to a MCP Server.

The HTTP transport allows the MCP Client and Server to be in the same environment or distributed over the Internet.

The connection over HTTP will look similar to this:

mcp.run(
        transport="http",
        host="0.0.0.0",
        port=port,
    )

Running the Python Code

First- switch the directory with the Python MCP sample code. Then, run the release version on the local system:

cd ~/gemini-cli-aws/mcp-https-python-ecs-express
make release

If everything is running correctly — you will see the FASTMCP banner:


                 ╭──────────────────────────────────────────────────────────────────────────────╮                 
                 │ │                 
                 │ │                 
                 │ ▄▀▀ ▄▀█ █▀▀ ▀█▀ █▀▄▀█ █▀▀ █▀█ │                 
                 │ █▀ █▀█ ▄▄█ █ █ ▀ █ █▄▄ █▀▀ │                 
                 │ │                 
                 │ │                 
                 │ FastMCP 3.2.0 │                 
                 │ https://gofastmcp.com │                 
                 │ │                 
                 │ 🖥 Server: hello-world-server, 3.2.0 │                 
                 │ 🚀 Deploy free: https://horizon.prefect.io │                 
                 │ │                 
                 ╰──────────────────────────────────────────────────────────────────────────────╯                 

[04/04/26 19:10:50] INFO Starting MCP server 'hello-world-server' with transport 'http' on transport.py:299
                             http://0.0.0.0:8080/mcp                                                              
INFO: Started server process [27502]
INFO: Waiting for application startup.
{"message": "StreamableHTTP session manager started"}
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8080 (Press CTRL+C to quit)

At this point — hit Cntrl-C to exit the local HTTP server.

Deploy to ECS Express

Now that the server has been tested locally- start the remote deployment to ECS Express.

Use the deploy target:

 > make deploy
✦ I will execute the full deployment cycle, which includes building the Docker image, pushing it to ECR, and
  deploying to ECS Express Mode.

Then check the status:

 > make status
✦ I will check the status of the ECS Express Mode service.

╭────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ ✓ Shell make status [in /home/xbill/gemini-cli-aws/mcp-ecsexpress-python-aws] │
│ │
│ Checking AWS ECS service status for mcp-express-python-aws... │
│ Service: mcp-express-python-aws │
│ Status: ACTIVE │
│ Endpoint: mc-8d69e7dfa87344be98fdb3f7a8fbbbba.ecs.us-east-1.on.aws │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
✦ The ECS Express Mode service mcp-express-python-aws is ACTIVE and available at:
  mc-8d69e7dfa87344be98fdb3f7a8fbbbba.ecs.us-east-1.on.aws

And the endpoint:

  Based on the make status output earlier, the actual endpoint is:
  mc-8d69e7dfa87344be98fdb3f7a8fbbbba.ecs.us-east-1.on.aws

Gemini CLI settings.json

The default Gemini CLI settings.json has an entry for the Python source:

{
  "mcpServers": {
    "aws-ecs-express-python": {
      "httpUrl": "https://mc-8d69e7dfa87344be98fdb3f7a8fbbbba.ecs.us-east-1.on.aws/mcp"
    }
  }
}

Validation with Gemini CLI

Leave the HTTP server running and open a new terminal.

Gemini CLI is restarted and the MCP connection over HTTP to the Python Code is validated. The full Gemini CLI Session will start:


xbill@penguin:~/gemini-cli-aws/mcp-ecsexpress-python-aws$ gemini

 ▝▜▄ Gemini CLI v0.36.0
   ▝▜▄
  ▗▟▀ Signed in with Google: xbill@glitnir.com /auth
 ▝▀ Plan: Gemini Code Assist Standard /upgrade

 > /mcp list
Configured MCP servers:

🟢 aws-ecs-express-python - Ready (1 tool)
  Tools:
  - mcp_aws-ecs-express-python_greet

And you can then connect to the MCP Server over HTTP:

 > greet ECS Express!
✦ I will call the greet tool with "ECS Express!".

╭────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ ✓ greet (aws-ecs-express-python MCP Server) {"param":"ECS Express!"} │
│ │
│ ECS Express! │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
✦ The greet tool returned: ECS Express!

Validate Project Setup with Gemini CLI

Gemini CLI was used to summarize the project:

 > summarize the project deployment to Amazon ECS Express
✦ The deployment of the mcp-express-python-aws project to Amazon ECS Express Mode is a streamlined process that
  automates the container lifecycle from build to public endpoint.

  Deployment Workflow (make deploy)
   1. Containerization: Builds a Python 3.13-slim Docker image and pushes it to an Amazon ECR repository
      (mcp-server-repo).
   2. Identity & Access: Automatically provisions necessary IAM roles (ecsTaskExecutionRole and
      ecsInfrastructureRoleForExpressServices) if they do not already exist.
   3. Orchestration: Deploys the container to ECS Express Mode, which simplifies AWS Fargate deployments by
      automatically managing the Application Load Balancer, Security Groups, and Networking.

  Current Infrastructure State
   * Service Name: mcp-express-python-aws
   * Status: ACTIVE
   * Compute: AWS Fargate (Task size: 1 vCPU / 2GB RAM)
   * Scaling: Configured for 1 to 20 tasks, triggered at 60% average CPU usage.
   * Public Access: mc-8d69e7dfa87344be98fdb3f7a8fbbbba.ecs.us-east-1.on.aws (Port 8080)

  The service is currently running and reachable via its assigned ECS Express Gateway endpoint.

Summary

The strategy for using Python for MCP development with Gemini CLI and AWS was validated with a incremental step by step approach.

A minimal HTTP transport MCP Server was started from Python source code and validated with Gemini CLI running as a MCP client in the same local environment.

The HTTP server was then tested remotely from the AWS ECS Express instance.

This approach can be extended to more complex deployments using other MCP transports and Cloud based options.

Deploying ADK Agents on Azure Fabric

xbill — Mon, 06 Apr 2026 13:07:27 +0000

Leveraging the Google Agent Development Kit (ADK) and the underlying Gemini LLM to build Agents in the Python programming language deployed on the Azure Cloud platform.

What is Gemini CLI?

The Gemini CLI is an open-source, terminal-based AI agent from Google that allows developers to interact directly with Gemini models, such as Gemini 2.5 Pro, for coding, content creation, and workflow automation. It supports file operations, shell commands, and connects to external tools via the Model Context Protocol (MCP).

The full details on Gemini CLI are available here:

Build, debug & deploy with AI

Azure Fabric

Microsoft Fabric is a comprehensive, AI-powered cloud platform that unifies data engineering, data warehousing, data science, and business intelligence (using Power BI) into a single SaaS solution, built on the OneLake storage system. It simplifies data management by reducing reliance on disparate, siloed tools.

More details are here:

What is Microsoft Fabric - Microsoft Fabric

Why would I want Gemini CLI with Azure? Isn’t that a Google Thing?

Python Version Management

One of the downsides of the wide deployment of Python has been managing the language versions across platforms and maintaining a supported version.

The pyenv tool enables deploying consistent versions of Python:

GitHub - pyenv/pyenv: Simple Python version management

As of writing — the mainstream python version is 3.13. To validate your current Python:

python --version
Python 3.13.12

Gemini CLI

If not pre-installed you can download the Gemini CLI to interact with the source files and provide real-time assistance:

npm install -g @google/gemini-cli

Testing the Gemini CLI Environment

Once you have all the tools and the correct Node.js version in place- you can test the startup of Gemini CLI. You will need to authenticate with a Key or your Google Account:

azureuser@azure-new:~/gemini-cli-azure$ gemini

▝▜▄ Gemini CLI v0.33.1
    ▝▜▄
   ▗▟▀ Logged in with Google /auth
  ▝▀ Gemini Code Assist Standard /upgrade

Node Version Management

Gemini CLI needs a consistent, up to date version of Node. The nvm command can be used to get a standard Node environment:

GitHub - nvm-sh/nvm: Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions

To verify the Node version:

node --version
v25.9.0
nvm --version
0.40.3
npm --version
11.12.1

Agent Development Kit

The ADK can be installed from here:

Agent Development Kit (ADK)

Where do I start?

The strategy for starting Agent development is a incremental step by step approach.

The agents in the demo are based on the original code lab:

Create and deploy low code ADK (Agent Deployment Kit) agents using ADK Visual Builder | Google Codelabs

First, the basic development environment is setup with the required system variables, and a working Gemini CLI configuration.

Then, a minimal ADK Agent is built with the visual builder. Next — the entire solution is deployed to Google Cloud Run.

Setup the Basic Environment

At this point you should have a working Python environment and a working Gemini CLI installation. The next step is to clone the GitHub samples repository with support scripts:

cd ~
git clone https://github.com/xbill9/gemini-cli-azure

Then run init.sh from the cloned directory.

The script will attempt to determine your shell environment and set the correct variables:

source init.sh

If your session times out or you need to re-authenticate- you can run the set_env.sh script to reset your environment variables:

source set_env.sh

Variables like PROJECT_ID need to be setup for use in the various build scripts- so the set_env script can be used to reset the environment if you time-out.

Azure Login

All of the deployment is done with the az CLI tool- but it also needs an active login on the Azure environment. To check the version:

xbill@penguin:~/gemini-cli-azure/adkui-aci$ az --version
azure-cli 2.84.0
core 2.84.0
telemetry 1.1.0

Dependencies:
msal 1.35.0b1
azure-mgmt-resource 24.0.0
Python location '/opt/az/bin/python3'
Config directory '/home/xbill/.azure'
Extensions directory '/home/xbill/.azure/cliextensions'
Python (Linux) 3.13.11 (main, Feb 25 2026, 02:29:30) [GCC 12.2.0]
Legal docs and information: aka.ms/AzureCliLegal

Your CLI is up-to-date.

To login to Azure:

az login
You have logged into Microsoft Azure!
You can close this window, or we will redirect you to the Azure CLI documentation in 1 minute.

Verify The ADK Installation

To verify the setup, run the ADK CLI with Agent3:

xbill@penguin:~/gemini-cli-azure/adkui-fabric$ adk run Agent3
Log setup complete: /tmp/agents_log/agent.20260402_182048.log
To access latest log: tail -F /tmp/agents_log/agent.latest.log
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/utils/agent_loader.py:277: UserWarning: [EXPERIMENTAL] _load_from_yaml_config: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  if root_agent := self._load_from_yaml_config(actual_agent_name, agents_dir):
/home/xbill/.local/lib/python3.13/site-packages/google/adk/features/_feature_decorator.py:81: UserWarning: [EXPERIMENTAL] feature FeatureName.AGENT_CONFIG is enabled.
  check_feature_enabled()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/cli.py:204: UserWarning: [EXPERIMENTAL] InMemoryCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  credential_service = InMemoryCredentialService()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/auth/credential_service/in_memory_credential_service.py:33: UserWarning: [EXPERIMENTAL] BaseCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  super(). __init__ ()
Running agent studio_director, type exit to exit.



  
  
  Running the ADK Web Interface


First- start the ADK interface:



xbill@penguin:~/gemini-cli-azure/adkui-fabric$ adk web --host 0.0.0.0
2026-04-02 18:22:15,410 - INFO - service_factory.py:266 - Using in-memory memory service
2026-04-02 18:22:15,410 - INFO - local_storage.py:84 - Using per-agent session storage rooted at /home/xbill/gemini-cli-azure/adkui-fabric
2026-04-02 18:22:15,410 - INFO - local_storage.py:110 - Using file artifact service at /home/xbill/gemini-cli-azure/adkui-fabric/.adk/artifacts
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/fast_api.py:192: UserWarning: [EXPERIMENTAL] InMemoryCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  credential_service = InMemoryCredentialService()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/auth/credential_service/in_memory_credential_service.py:33: UserWarning: [EXPERIMENTAL] BaseCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  super(). __init__ ()
INFO: Started server process [27984]
INFO: Waiting for application startup.

+-----------------------------------------------------------------------------+
| ADK Web Server started |
| |
| For local testing, access at [http://0.0.0.0:8000](http://0.0.0.0:8000). |
+-----------------------------------------------------------------------------+

INFO: Application startup complete.
INFO: Uvicorn running on [http://0.0.0.0:8000](http://0.0.0.0:8000) (Press CTRL+C to quit)







  
  
  Connecting to the ADK


The ADK web interface is started with the place holder 0.0.0.0 address — which tells the service to listen on all interfaces. Depending on your local environment setup- you may need to use the actual local interface. There are 2 potential options:




http://0.0.0.0:8000

http://127.0.0.1:8000






This will bring up the ADK UI:




  
  
  Run a Sub Agent


Connect to the ADK and select “Agent3”:



The first interaction just verifies the connection to the agent:




what do you do?

I manage a linear pipeline of four ADK Sequential Agents to transform a user's seed idea into a fully rendered, responsive HTML5 comic book. If your request starts with "Create me a comic of...", I'll delegate the task to my sub-agent to begin the comic creation pipeline. Otherwise, I'll answer your question directly.






Give the Agent this prompt:




Create a comic of Bob and Clippy in a Battle Royal with Tux.






The sub agent will then generate the comic book images:




  
  
  Visual Build an Agent


To use the ADK visual builder- select the pencil Icon next to Agent 2. You can drill down into the Agent design:




  
  
  What did I just do?


The initial steps validate that the Python code and the ADK environment is functional and that the Agent code will run correctly. The next step is to deploy this environment to the Azure Fabric.


  
  
  Deploying to Azure Fabric


To start the deployment:




xbill@penguin:~/gemini-cli-azure/adkui-fabric$ make deploy
Ensuring Resource Group adk-rg-fabric exists...






You can validate the final result by checking the messages:




✦ The deployment to Azure Container Apps (Fabric integration) has completed successfully.

  Deployment Details:
   * Resource Group: adk-rg-fabric
   * ACR: adkacrpenguinv2
   * Container Apps Environment: adk-env-fabric
   * Container App: adk-comic-fabric
   * Endpoint: https://adk-comic-fabric.nicebush-151ee659.westus2.azurecontainerapps.io
   * Status: Running (Provisioning State: Succeeded)

  All environment variables (including API keys and project IDs) have been correctly applied to the Container App.
  You can now use this endpoint in your Microsoft Fabric Workload Manifest.






You can then get the endpoint:




xbill@penguin:~/gemini-cli-azure/adkui-fabric$ make endpoint
adk-comic-fabric.nicebush-151ee659.westus2.azurecontainerapps.io
xbill@penguin:~/gemini-cli-azure/adkui-fabric$ 







  
  
  Running the ADK Web Interface


Start a connection to the Azure Deployed ADK:



This will bring up the ADK UI.

Select the sub-agent “Agent3” and ask the Agent to create a comic about Bit:




Bit the Raccoon is a popular, unofficial community mascot for Microsoft Azure,
specifically representing the Cloud Advocate team and open-source initiatives. 
Created by Ashley McNamara, Bit is often depicted in tech-related, helpful 
scenarios to make complex cloud concepts more accessible and approachable 
to developers. X X +2 Role: Known as the "Developer Advocate Mascot," 
Bit is often used in presentations and content to represent the "open" 
aspect of Azure. Context: While Azure has an official anime-style mascot 
in Japan (Claudia Madobe), Bit is beloved within the broader developer 
community as a fun, tech-savvy character.  

Appearance: Bit is usually styled as an adorable, helpful raccoon, 
sometimes accompanied by other community mascots like "Brady the Ninja Sloth."
GitHub GitHub Bit is a favorite for stickers and frequently appears 
on social media to highlight new Azure features.






The Agent will get to work:




  
  
  Run the Viewer Agent


Connect to the ADK and select “Agent4”:




I've exported the comic as ADK artifacts for you!

You can now view the comic directly within the ADK environment. Please check the 'Artifacts' pane or tab in your ADK Visual Builder / Studio UI.

You'll find the following:

view_full_comic.html: This is a complete, self-contained HTML file with embedded images for the full comic experience.
comic_preview.md: A markdown summary of the comic.
Individual panel images (e.g., panel_panel_1.png).
If you are working in a local terminal, you can still view the original local files by opening output/comic.html in your browser or by running ./comic.sh. However, the ADK artifacts are the best way to view it while hosted!






The Agent will generate the artifacts:



The Comic can then be viewed:




  
  
  Summary


The Agent Development Kit was used to visually define a basic agent and added a tool that calls the ImageGen Gemini Model. This Agent was tested locally with the CLI and then with the ADK web tool. Then, the Agent was packaged and deployed to Azure Fabric. Several sample ADK agents were run directly from the deployment in Azure. This approach validates that cross cloud tools can be used — even with more complex agents.

Deploying ADK Agents on Azure Kubernates Service (AKS)

xbill — Sat, 04 Apr 2026 22:47:18 +0000

Leveraging the Google Agent Development Kit (ADK) and the underlying Gemini LLM to build Agents in the Python programming language deployed on the Azure Cloud platform.

What is Gemini CLI?

The full details on Gemini CLI are available here:

Build, debug & deploy with AI

Azure Kubernates Service

More details are available here:

https://azure.microsoft.com/en-us/products/kubernetes-service

Isn’t that Overkill? A whole Cluster Just for some Agents?!

Why would I want Gemini CLI with Azure? Isn’t that a Google Thing?

Python Version Management

One of the downsides of the wide deployment of Python has been managing the language versions across platforms and maintaining a supported version.

The pyenv tool enables deploying consistent versions of Python:

GitHub - pyenv/pyenv: Simple Python version management

As of writing — the mainstream python version is 3.13. To validate your current Python:

xbill@penguin:~/gemini-cli-azure/adkui-aks$ python --version
Python 3.13.12

Gemini CLI

If not pre-installed you can download the Gemini CLI to interact with the source files and provide real-time assistance:

npm install -g @google/gemini-cli

Testing the Gemini CLI Environment

Once you have all the tools and the correct Node.js version in place- you can test the startup of Gemini CLI. You will need to authenticate with a Key or your Google Account:

gemini

▝▜▄ Gemini CLI v0.33.1
    ▝▜▄
   ▗▟▀ Logged in with Google /auth
  ▝▀ Gemini Code Assist Standard /upgrade

Node Version Management

Gemini CLI needs a consistent, up to date version of Node. The nvm command can be used to get a standard Node environment:

GitHub - nvm-sh/nvm: Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions

To verify the Node version:

xbill@penguin:~/gemini-cli-azure$ nvm --version
0.40.3
xbill@penguin:~/gemini-cli-azure$ node --version
v25.9.0
xbill@penguin:~/gemini-cli-azure$ npm --version
11.12.1

Agent Development Kit

The ADK can be installed from here:

Agent Development Kit (ADK)

Where do I start?

The strategy for starting Agent development is a incremental step by step approach.

The agents in the demo are based on the original code lab:

Create and deploy low code ADK (Agent Deployment Kit) agents using ADK Visual Builder | Google Codelabs

First, the basic development environment is setup with the required system variables, and a working Gemini CLI configuration.

Then, a minimal ADK Agent is built with the visual builder. Next — the entire solution is deployed to Google Cloud Run.

Setup the Basic Environment

At this point you should have a working Python environment and a working Gemini CLI installation. The next step is to clone the GitHub samples repository with support scripts:

cd ~
git clone https://github.com/xbill9/gemini-cli-azure

Then run init.sh from the cloned directory.

The script will attempt to determine your shell environment and set the correct variables:

source init.sh

If your session times out or you need to re-authenticate- you can run the set_env.sh script to reset your environment variables:

source set_env.sh

Variables like PROJECT_ID need to be setup for use in the various build scripts- so the set_env script can be used to reset the environment if you time-out.

Azure Login

All of the deployment is done with the az CLI tool- but it also needs an active login on the Azure environment. To check the version:

xbill@penguin:~/gemini-cli-azure$ az --version
azure-cli 2.84.0

core 2.84.0
telemetry 1.1.0

Dependencies:
msal 1.35.0b1
azure-mgmt-resource 24.0.0

Python location '/opt/az/bin/python3'
Config directory '/home/xbill/.azure'
Extensions directory '/home/xbill/.azure/cliextensions'

Python (Linux) 3.13.11 (main, Feb 25 2026, 02:29:30) [GCC 12.2.0]

Legal docs and information: aka.ms/AzureCliLegal

Unable to check if your CLI is up-to-date. Check your internet connection.

To login to Azure:

az login

Then you will see a message after successful authentication:

You have logged into Microsoft Azure!
You can close this window, or we will redirect you to the Azure CLI documentation in 1 minute.

Verify The ADK Installation

To verify the setup, run the ADK CLI with Agent3:

xbill@penguin:~/gemini-cli-azure/adkui-aks$ adk run Agent3
Log setup complete: /tmp/agents_log/agent.20260402_173311.log
To access latest log: tail -F /tmp/agents_log/agent.latest.log
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/utils/agent_loader.py:277: UserWarning: [EXPERIMENTAL] _load_from_yaml_config: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  if root_agent := self._load_from_yaml_config(actual_agent_name, agents_dir):
/home/xbill/.local/lib/python3.13/site-packages/google/adk/features/_feature_decorator.py:81: UserWarning: [EXPERIMENTAL] feature FeatureName.AGENT_CONFIG is enabled.
  check_feature_enabled()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/cli.py:204: UserWarning: [EXPERIMENTAL] InMemoryCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  credential_service = InMemoryCredentialService()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/auth/credential_service/in_memory_credential_service.py:33: UserWarning: [EXPERIMENTAL] BaseCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  super(). __init__ ()
Running agent studio_director, type exit to exit.



  
  
  Running the ADK Web Interface


First- start the ADK interface:



xbill@penguin:~/gemini-cli-azure/adkui-aks$ adk web --host 0.0.0.0
2026-04-02 17:33:54,016 - INFO - service_factory.py:266 - Using in-memory memory service
2026-04-02 17:33:54,016 - INFO - local_storage.py:84 - Using per-agent session storage rooted at /home/xbill/gemini-cli-azure/adkui-aks
2026-04-02 17:33:54,016 - INFO - local_storage.py:110 - Using file artifact service at /home/xbill/gemini-cli-azure/adkui-aks/.adk/artifacts
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/fast_api.py:192: UserWarning: [EXPERIMENTAL] InMemoryCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  credential_service = InMemoryCredentialService()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/auth/credential_service/in_memory_credential_service.py:33: UserWarning: [EXPERIMENTAL] BaseCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  super(). __init__ ()
INFO: Started server process [23049]
INFO: Waiting for application startup.

+-----------------------------------------------------------------------------+
| ADK Web Server started |
| |
| For local testing, access at http://0.0.0.0:8000. |
+-----------------------------------------------------------------------------+

INFO: Application startup complete.







  
  
  Connecting to the ADK


The ADK web interface is started with the place holder 0.0.0.0 address — which tells the service to listen on all interfaces. Depending on your local environment setup- you may need to use the actual local interface. There are 2 potential options:




http://0.0.0.0:8000

http://127.0.0.1:8000






This will bring up the ADK UI:




  
  
  Run a Sub Agent


Connect to the ADK and select “Agent3”:



The first interaction just verifies the connection to the agent:




what do you do?

I manage a linear pipeline of four ADK Sequential Agents to transform a user's seed idea into a fully rendered, responsive HTML5 comic book. If your request starts with "Create me a comic of...", I'll delegate the task to my sub-agent to begin the comic creation pipeline. Otherwise, I'll answer your question directly.






Give the Agent this prompt:




Create a comic of Bob and Clippy in a Battle Royal with Tux.






The sub agent will then generate the comic book images:




  
  
  Visual Build an Agent


To use the ADK visual builder- select the pencil Icon next to Agent 2. You can drill down into the Agent design:




  
  
  What did I just do?


The initial steps validate that the Python code and the ADK environment is functional and that the Agent code will run correctly. The next step is to deploy this environment to the Azure Kubernates Service.


  
  
  Deploying to Azure Kubernates Service


To start the deployment:




✦ I will now run the deploy-aks.sh script 






You can validate the final result by checking the messages:




✦ The ADK application has been successfully deployed to AKS.

  Deployment Summary:
   - Deployment Status: adk-app is Ready (1/1).
   - Pod Status: adk-app-dcf865487-nbfrj is Running.
   - Service Endpoint: http://4.155.123.101 (http://4.155.123.101)

  The application is now accessible at the public IP listed above. Let me know if you need anything else!






You can then get the endpoint:




✦ The public endpoint for the ADK application is:
  4.155.123.101






The service will be visible in the Azure console:




  
  
  Running the ADK Web Interface


Start a connection to the Azure Deployed ADK.

This will bring up the ADK UI:



Select the sub-agent “Agent3” and ask the Agent to create a comic about Bit:




Bit the Raccoon is a popular, unofficial community mascot for Microsoft Azure,
specifically representing the Cloud Advocate team and open-source initiatives. 
Created by Ashley McNamara, Bit is often depicted in tech-related, helpful 
scenarios to make complex cloud concepts more accessible and approachable 
to developers. X X +2 Role: Known as the "Developer Advocate Mascot," 
Bit is often used in presentations and content to represent the "open" 
aspect of Azure. Context: While Azure has an official anime-style mascot 
in Japan (Claudia Madobe), Bit is beloved within the broader developer 
community as a fun, tech-savvy character.

Appearance: Bit is usually styled as an adorable, helpful raccoon, 
sometimes accompanied by other community mascots like "Brady the Ninja Sloth."
GitHub GitHub Bit is a favorite for stickers and frequently appears 
on social media to highlight new Azure features.

Create a story about bit






The Agent will get to work:




  
  
  Run the Viewer Agent


Connect to the ADK and select “Agent4”:




I've exported the comic as ADK artifacts for you!

You can now view the comic directly within the ADK environment. Please check the 'Artifacts' pane or tab in your ADK Visual Builder / Studio UI.

You'll find the following:

view_full_comic.html: This is a complete, self-contained HTML file with embedded images for the full comic experience.
comic_preview.md: A markdown summary of the comic.
Individual panel images (e.g., panel_panel_1.png).
If you are working in a local terminal, you can still view the original local files by opening output/comic.html in your browser or by running ./comic.sh. However, the ADK artifacts are the best way to view it while hosted!






The Agent will generate the artifacts:



The Comic can then be viewed:




  
  
  Summary


The Agent Development Kit was used to visually define a basic agent and added a tool that calls the Image Gemini Model. This Agent was tested locally with the CLI and then with the ADK web tool. Then, the Agent was packaged and deployed to the Azure Kubernates Service. Several sample ADK agents were run directly from the deployment in Azure. This approach validates that cross cloud tools can be used — even with more complex agents.

Building a Multimodal Cross Cloud Live Agent with ADK, Azure ACI, and Gemini CLI

xbill — Sat, 04 Apr 2026 22:45:43 +0000

Leveraging the Google Agent Development Kit (ADK) and the underlying Gemini LLM to build cross cloud apps with the Python programming language deployed to the Azure ACI service.

Aren’t There a Billion Python MCP Demos?

Yes there are.

What Is Python?

Python is an interpreted language that allows for rapid development and testing and has deep libraries for working with ML and AI:

Welcome to Python.org

Python Version Management

One of the downsides of the wide deployment of Python has been managing the language versions across platforms and maintaining a supported version.

The pyenv tool enables deploying consistent versions of Python:

GitHub - pyenv/pyenv: Simple Python version management

As of writing — the mainstream python version is 3.13. To validate your current Python:

admin@ip-172-31-70-211:~/gemini-cli-azure$ python --version
Python 3.13.12

Azure Container Instances

Azure Container Instances (ACI) is a serverless, managed service that allows you to run Docker containers in the cloud without managing virtual machines. It is ideal for rapid deployment, bursting, and simple, isolated applications, offering per-second billing and quick startup times. ACI supports Linux and Windows containers, with options for volume mounting and GPU resources.

More details are available here:

https://azure.microsoft.com/en-us/products/container-instances

Why would I want Gemini CLI with Azure? Isn’t that a Google Thing?

Gemini CLI

If not pre-installed you can download the Gemini CLI to interact with the source files and provide real-time assistance:

npm install -g @google/gemini-cli

Testing the Gemini CLI Environment

Once you have all the tools and the correct Node.js version in place- you can test the startup of Gemini CLI. You will need to authenticate with a Key or your Google Account:

gemini

admin@ip-172-31-70-211:~/gemini-cli-azure$ gemini

▝▜▄ Gemini CLI v0.33.1
    ▝▜▄
   ▗▟▀ Logged in with Google /auth
  ▝▀ Gemini Code Assist Standard /upgrade

? for shortcuts 
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 shift+tab to accept edits 3 GEMINI.md files | 1 MCP server
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 > Type your message or @path/to/file
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 ~/.../gemini-cli-azure (main*) no sandbox (see /docs) /model Auto (Gemini 3) | 239.8 MB

Node Version Management

Gemini CLI needs a consistent, up to date version of Node. The nvm command can be used to get a standard Node environment:

GitHub - nvm-sh/nvm: Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions

Docker Version Management

The Azure CLI tools need current version of Docker. If your environment does not provide a recent docker tool- the Docker Version Manager can be used to downlaod the latest supported Docker:

Install

Azure CLI

The Azure CLI provides a command line tool to directly access Azure services from your current environment. Full details on the CLI are available here:

Azure Command-Line Interface (CLI) documentation

Agent Development Kit

The ADK can be installed from here:

Agent Development Kit (ADK)

This seems like a lot of Configuration!

Getting the key tools in place is the first step to working across Cloud environments. For a deeper dive- a project with a similar setup can be found here:

Deploying ADK Agents on Azure ACI

Where do I start?

The strategy for starting multimodal real time cross cloud agent development is a incremental step by step approach.

The agents in the demo are based on the original code lab:

Way Back Home - Building an ADK Bi-Directional Streaming Agent | Google Codelabs

First, the basic development environment is setup with the required system variables, and a working Gemini CLI configuration.

Then, a minimal ADK Agent is built with the visual builder. Next — the entire solution is deployed to Google Cloud Run.

Setup the Basic Environment

The next step is to clone the GitHub repository to your local environment:

cd ~
git clone https://github.com/xbill9/gemini-cli-azure
cd level_3-aci

Then run init.sh from the cloned directory.

The script will attempt to determine your shell environment and set the correct variables:

source init.sh

If your session times out or you need to re-authenticate- you can run the set_env.sh script to reset your environment variables:

source set_env.sh

Variables like PROJECT_ID need to be setup for use in the various build scripts- so the set_env script can be used to reset the environment if you time-out.

Verify The ADK Installation

To verify the setup, run the ADK CLI locally with Agent1:

 /home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/cli.py:204: UserWarning: [EXPERIMENTAL] InM │
│ emoryCredentialService: This feature is experimental and may change or be removed in future versions witho │
│ ut notice. It may introduce breaking changes at any time. │
│ credential_service = InMemoryCredentialService() │
│ /home/xbill/.local/lib/python3.13/site-packages/google/adk/auth/credential_service/in_memory_credential_se │
│ rvice.py:33: UserWarning: [EXPERIMENTAL] BaseCredentialService: This feature is experimental and may chang │
│ e or be removed in future versions without notice. It may introduce breaking changes at any time. │
│ super(). __init__ () │
│ Running agent biometric_agent, type exit to exit. │
│ [user]:

Deploying to Azure Container Instances

The first step is to refresh the Azure credentials in the current build environment:

xbill@penguin:~/gemini-cli-azure/level_3-appservice$ az login

Run the deploy version on the local system:

 > make deploy
✦ I will execute make deploy to initiate the deployment process, which will build the Docker image and deploy it
  to Azure Container Instance using the updated gemini-2.5-flash-native-audio-latest model. 0.0s 0.0s

You can validate the final result by checking the messages:

✦ The application has been successfully deployed to Azure Container Instance using the
  gemini-2.5-flash-native-audio-latest model.

  Deployment Details:
   - Model Used: gemini-2.5-flash-native-audio-latest (Live/Bidirectional Audio support)
   - Deployment URL: http://biometric-scout-aci.canadaeast.azurecontainer.io:8080
     (http://biometric-scout-aci.canadaeast.azurecontainer.io:8080)
   - Azure Resource Group: level_3-aci
   - Azure Container Registry: biometricacrpenguinv2.azurecr.io

  All configuration files (.env, set_env.sh, init.sh, deploy-azure.sh, and agent.py) have been synchronized to use
  the 2.5 Flash live model as requested.

Once the container is deployed- you can then get the endpoint:

✦ The Azure Container Instance is currently Running.

  Service Details:
   - Status: Running
   - URL: http://biometric-scout-aci.canadaeast.azurecontainer.io:8080
     (http://biometric-scout-aci.canadaeast.azurecontainer.io:8080)
   - Model: gemini-2.5-flash-native-audio-latest (Native Audio Enabled)

  Your deployment using the 2.5 Flash live model is fully operational.

The service will be visible in the Azure console:

Running the Web Interface

Start a connection to the deployed app:

http://biometric-scout-aci.canadaeast.azurecontainer.io:8080

Note: ACI does not deploy to HTTPS by default. For basic HTTP deployment- Allowlist the ACI website as a trusted site.

Then connect to the app :

Then use the Live model to process audio and video:

Finally — complete the sequence:

Summary

The Agent Development Kit was used to enable a multi-modal agent using the Gemini Live Model. This Agent was tested locally with the CLI and then deployed to Azure Container Instance. This approach validates that cross cloud tools can be used — even with more complex agents.

Building a Multimodal Cross Cloud Live Agent with ADK, Azure ACA, and Gemini CLI

xbill — Fri, 03 Apr 2026 19:36:29 +0000

Leveraging the Google Agent Development Kit (ADK) and the underlying Gemini LLM to build cross cloud apps with the Python programming language deployed to the Azure Container App service.

Aren’t There a Billion Python MCP Demos?

Yes there are.

What Is Python?

Python is an interpreted language that allows for rapid development and testing and has deep libraries for working with ML and AI:

Welcome to Python.org

Python Version Management

One of the downsides of the wide deployment of Python has been managing the language versions across platforms and maintaining a supported version.

The pyenv tool enables deploying consistent versions of Python:

GitHub - pyenv/pyenv: Simple Python version management

As of writing — the mainstream python version is 3.13. To validate your current Python:

admin@ip-172-31-70-211:~/gemini-cli-azure$ python --version
Python 3.13.12

Azure Container App Service

Azure Container Apps (ACA) is a fully managed, serverless platform designed for running containerized applications and microservices without managing underlying infrastructure. Built on Azure Kubernetes Service (AKS), it offers built-in autoscaling (including to zero), traffic splitting for blue/green deployments, and Dapr integration, making it ideal for event-driven, API, and background processing workloads.

https://azure.microsoft.com/en-us/products/container-apps

Why would I want Gemini CLI with Azure? Isn’t that a Google Thing?

Azure Container App Configuration

To configure your Azure Service with the base system tools- this article provides a reference:

MCP Development with Python, and Azure Container Apps

Gemini CLI

If not pre-installed you can download the Gemini CLI to interact with the source files and provide real-time assistance:

npm install -g @google/gemini-cli

Testing the Gemini CLI Environment

Once you have all the tools and the correct Node.js version in place- you can test the startup of Gemini CLI. You will need to authenticate with a Key or your Google Account:

gemini

admin@ip-172-31-70-211:~/gemini-cli-azure$ gemini

▝▜▄ Gemini CLI v0.33.1
    ▝▜▄
   ▗▟▀ Logged in with Google /auth
  ▝▀ Gemini Code Assist Standard /upgrade

? for shortcuts 
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 shift+tab to accept edits 3 GEMINI.md files | 1 MCP server
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 > Type your message or @path/to/file
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 ~/.../gemini-cli-azure (main*) no sandbox (see /docs) /model Auto (Gemini 3) | 239.8 MB

Node Version Management

Gemini CLI needs a consistent, up to date version of Node. The nvm command can be used to get a standard Node environment:

GitHub - nvm-sh/nvm: Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions

Docker Version Management

The Azure CLI tools need current version of Docker. If your environment does not provide a recent docker tool- the Docker Version Manager can be used to downlaod the latest supported Docker:

Install

Azure CLI

The Azure CLI provides a command line tool to directly access Azure services from your current environment. Full details on the CLI are available here:

Azure Command-Line Interface (CLI) documentation

Agent Development Kit

The ADK can be installed from here:

Agent Development Kit (ADK)

This seems like a lot of Configuration!

Getting the key tools in place is the first step to working across Cloud environments. For a deeper dive- a project with a similar setup can be found here:

Deploying ADK Agents on Azure ACA (Azure Container Apps)

Where do I start?

The strategy for starting multimodal real time cross cloud agent development is a incremental step by step approach.

The agents in the demo are based on the original code lab:

Way Back Home - Building an ADK Bi-Directional Streaming Agent | Google Codelabs

First, the basic development environment is setup with the required system variables, and a working Gemini CLI configuration.

Then, a minimal ADK Agent is built with the visual builder. Next — the entire solution is deployed to Google Cloud Run.

Setup the Basic Environment

The next step is to clone the GitHub repository to your local environment:

cd ~
git clone https://github.com/xbill9/gemini-cli-azure
cd level_3-aca

Then run init.sh from the cloned directory.

The script will attempt to determine your shell environment and set the correct variables:

source init.sh

If your session times out or you need to re-authenticate- you can run the set_env.sh script to reset your environment variables:

source set_env.sh

Variables like PROJECT_ID need to be setup for use in the various build scripts- so the set_env script can be used to reset the environment if you time-out.

Verify The ADK Installation

To verify the setup, run the ADK CLI locally with Agent1:

xbill@penguin:~/gemini-cli-azure/level_3-aca/backend/app$ adk run biometric_agent
Log setup complete: /tmp/agents_log/agent.20260402_203152.log
To access latest log: tail -F /tmp/agents_log/agent.latest.log
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/cli.py:204: UserWarning: [EXPERIMENTAL] InMemoryCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  credential_service = InMemoryCredentialService()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/auth/credential_service/in_memory_credential_service.py:33: UserWarning: [EXPERIMENTAL] BaseCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  super(). __init__ ()
Running agent biometric_agent, type exit to exit.



  
  
  Deploying to Azure Container Apps


The first step is to refresh the Azure credentials in the current build environment:



xbill@penguin:~/gemini-cli-azure/level_3-aca$ az login






Run the deploy version on the local system:




 > make deploy
✦ I will execute make deploy, which builds the Docker image and deploys it to Azure Container Apps, creating the
  necessary resource groups, container registry, and environment as defined in deploy.sh. 0.0s 0.0s






You can validate the final result by checking the messages:




✦ The deployment to Azure Container Apps was successful. You can access the application at the following URL:

  https://biometric-scout-app.agreeablebeach-34887419.canadaeast.azurecontainerapps.io






Once the container is deployed- you can then get the endpoint:




✦ The endpoint for your Azure Container App is:
  biometric-scout-app.agreeablebeach-34887419.canadaeast.azurecontainerapps.io






The service will be visible in the Azure console:




  
  
  Running the Web Interface


Start a connection to the deployed app:




biometric-scout-app.agreeablebeach-34887419.canadaeast.azurecontainerapps.io






Then connect to the app :



Then use the Live model to process audio and video:



Finally — complete the sequence:




  
  
  Summary


The Agent Development Kit was used to enable a multi-modal agent using the Gemini Live Model. This Agent was tested locally with the CLI and then deployed to Azure Container App services. This approach validates that cross cloud tools can be used — even with more complex agents.

MCP Development with Python, and Azure Kubernates Service (AKS)

xbill — Fri, 03 Apr 2026 18:59:56 +0000

Leveraging Gemini CLI and the underlying Gemini LLM to build Model Context Protocol (MCP) AI applications with Python with a local development environment deployed to the Azure Kubernates Service.

What is Gemini CLI?

The full details on Gemini CLI are available here:

Build, debug & deploy with AI

Azure Kubernates Service

More details are available here:

https://azure.microsoft.com/en-us/products/kubernetes-service

Isn’t that Overkill? A whole Cluster Just for some MCP?!

An entire cluster is a large deployment for just a basic MCP server. The goal was to validate that MCP servers could be deployed — and that opens the door for more complex deployments that can take advantage of the full services in the cluster.

Why would I want Gemini CLI with Azure? Isn’t that a Google Thing?

Node Version Management

Gemini CLI needs a consistent, up to date version of Node. The nvm command can be used to get a standard Node environment:

GitHub - nvm-sh/nvm: Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions

Gemini CLI Installation

You can then download the Gemini CLI :

npm install -g @google/gemini-cli

Testing the Gemini CLI Environment

Once you have all the tools and the correct Node.js version in place- you can test the startup of Gemini CLI. You will need to authenticate with a Key or your Google Account:

gemini

Authentication

Several authentication options are available. To use an existing Code Assist licence — authenticate with a Google Account:

> /auth                                                                                                                                                        
▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄
╭──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ │
│ ? Get started │
│ │
│ How would you like to authenticate for this project? │
│ │
│ ● 1. Login with Google │
│ 2. Use Gemini API Key │
│ 3. Vertex AI │
│ │
│ (Use Enter to select) │
│ │
│ Terms of Services and Privacy Notice for Gemini CLI │
│ │
│ https://geminicli.com/docs/resources/tos-privacy/ │
│ │
╰──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

Then set the GOOGLE_CLOUD_PROJECT to a valid project setup on the Google Cloud console:

~ $ export GOOGLE_CLOUD_PROJECT=comglitn
~ $

Other options include Google Cloud API Key that can be generated directly from the Google Cloud Console.

Installing Google Cloud Tools

To simplify working with Google Cloud — install the Google Cloud Tools:

https://docs.cloud.google.com/sdk/docs/install-sdk

Once the installation is completed — you can verify the setup:

william@Azure:~$ gcloud auth list
  Credentialed Accounts
ACTIVE ACCOUNT
* xbill@glitnir.com

Installing Azure Customized GEMINI.md

A sample GitHub repo contains tools for working with Gemini CLI on Azure. This repo is available here:

git clone https://gitHub.com/xbill9/gemini-cli-azure

A sample GEMINI.md customized for the Azure environment is provided in the repo:

This is a multi linux git repo hosted at:
github.com/xbill9/gemini-cli-azure

You are a cross platform developer working with 
Microsoft Azure and Google Cloud

You can use the Azure CLI :
https://learn.microsoft.com/en-us/cli/azure/install-azure-cli
https://learn.microsoft.com/en-us/cli/azure/
https://learn.microsoft.com/en-us/cli/azure/reference
https://learn.microsoft.com/en-us/cli/azure/install-azure-cli-linux?view=azure-cli-latest&pivots=apt
## Azure CLI Tools
You can use the Azure CLI to manage resources across Azure Storage, Virtual Machines, and other services.
- **List Resource Groups** : `az group list -o table`
- **List Storage Accounts** : `az storage account list -o table`
- **List Virtual Machines** : `az vm list -d -o table`
### Azure Update Script
- `azure-update`: This script is specifically for Azure Linux environments. It updates all packages and ensures necessary libraries are installed.
## Automation Scripts
This repository contains scripts for updating various Linux environments and tools:
- `linux-update`: Detects OS (Debian/Ubuntu/Azure Linux) and runs the corresponding update scripts.
- `azure-update`: Updates Azure Linux packages and installs necessary dependencies.
- `debian-update`: Updates Debian/Ubuntu packages and installs `git`.
- `gemini-update`: Updates the `@google/gemini-cli` via npm and checks versions of Node.js and Gemini.
- `nvm-update`: Installs NVM (Node Version Manager) and Node.js version 25.

Python MCP Documentation

The official GitHub Repo provides samples and documentation for getting started:

GitHub - modelcontextprotocol/python-sdk: The official Python SDK for Model Context Protocol servers and clients

The most common MCP Python deployment path uses the FASTMCP library:

Welcome to FastMCP - FastMCP

Where do I start?

The strategy for starting MCP development is a incremental step by step approach.

First, the basic development environment is setup with the required system variables, and a working Gemini CLI configuration.

Then, a minimal Hello World Style Python MCP Server was built with stdio transport. This server was validated with Gemini CLI in the local environment.

This current setup validates the connection from Gemini CLI to the local process via MCP. The MCP client (Gemini CLI) and the Python MCP server both run in the same local environment.

Next- the basic MCP server is extended with Gemini CLI to add several new tools in standard Python code.

Setup the Basic Environment

At this point you should have a working Python interpreter and a working Gemini CLI installation. The next step is to clone the GitHub samples repository with support scripts:

cd ~
git clone https://github.com/xbill9/gemini-cli-azure

Then run init.sh from the cloned directory.

The script will attempt to determine your shell environment and set the correct variables:

cd gemini-cli-azure
source init.sh

If your session times out or you need to re-authenticate- you can run the set_env.sh script to reset your environment variables:

cd gemini-cli-azure
source set_env.sh

Variables like PROJECT_ID need to be setup for use in the various build scripts- so the set_env script can be used to reset the environment if you time-out.

Hello World with HTTP Transport

One of the key features that the standard MCP libraries provide is abstracting various transport methods.

The high level MCP tool implementation is the same no matter what low level transport channel/method that the MCP Client uses to connect to a MCP Server.

The HTTP transport allows the MCP Client and Server to be in the same environment or distributed over the Internet.

The connection over HTTP will look similar to this:

mcp.run(
        transport="http",
        host="0.0.0.0",
        port=port,
    )

Running the Python Code

First- switch the directory with the Python MCP sample code:

CD mcp-aks-python-azure
make install
make run

You can validate the final result by checking the messages:

[04/02/26 16:02:19] INFO Starting MCP server 'hello-world-server' with transport 'http' on transport.py:299
                             http://0.0.0.0:8080/mcp                                                              
INFO: Started server process [16625]
INFO: Waiting for application startup.
{"message": "StreamableHTTP session manager started"}
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8080 (Press CTRL+C to quit)

Once you have validated the server can run locally — exit with control-c.

Then run a deployment to Azure Kubernates Service:

 > make deploy
✦ I'll check if the AKS cluster exists before proceeding with the deployment.

✦ The make deploy command was successful. The Python-based MCP server is now running on Azure Kubernetes Service
  (AKS).

  Deployment Details:
   - ACR Image: mcpacrpenguinv2.azurecr.io/mcp-server-image:921e6f8
   - AKS Cluster: mcp-aks-penguin
   - AKS Service IP: 20.99.165.219
   - Pod Status: Running (1/1 READY)

  You can now access the MCP server at http://20.99.165.219.

Getting deployment status:

✦ The current status of your Azure resources is as follows:

   - AKS Cluster (mcp-aks-penguin): Succeeded, version 1.34.
   - Kubernetes Service (mcp-server-service): LoadBalancer type, with External IP 20.99.165.219.
   - Kubernetes Pods: The mcp-server pod is Running and healthy (1/1 READY).

And the current endpoint:

│ │AKS Endpoint: │
│ 20.99.165.219

Gemini CLI settings.json

The default Gemini CLI settings.json has an entry for the Python source:

   1 {
   2 "mcpServers": {
   3 "azure-aks-python": {
   4 "httpUrl": "http://20.99.165.219/mcp"
   5 }
   6 }
   7 }

Validation with Gemini CLI

Leaver the MCP server window running. Start a new shell. Gemini CLI is restarted and the MCP connection over HTTP to the Python Code is validated, The full Gemini CLI Session will start:

  ▝▜▄ Gemini CLI v0.34.0
    ▝▜▄
   ▗▟▀ Signed in with Google: xbill@glitnir.com /auth
  ▝▀ Plan: Gemini Code Assist Standard /upgrade

> /mcp list

Configured MCP servers:

🟢 azure-aks-python - Ready (1 tool)
  Tools:
  - mcp_azure-aks-python_greet

🟢 DeveloperKnowledge (from DeveloperKnowledge) - Ready (2 tools)
  Tools:
  - mcp_DeveloperKnowledge_get_documents
  - mcp_DeveloperKnowledge_search_documents

And you can then connect to the MCP Server over HTTP:

 > greet AKS!
✦ I will now greet AKS! by calling the mcp_azure-aks-python_greet tool.

╭────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ ✓ greet (azure-aks-python MCP Server) {"param":"AKS!"} │
│ │
│ AKS! │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
✦ AKS!

Project Review

Finally — you can use Gemini CLI to review the project:

✦ I have conducted a comprehensive code review of the project. Overall, the implementation is clean, follows
  cloud-native best practices, and is well-configured for AKS. Here are my specific findings and recommendations:

  1. Architecture & Implementation (main.py)
   - FastMCP: Excellent use of the FastMCP library, which simplifies server creation and tool definition.
   - Health Check: The /health custom route is correctly implemented and integrated with Kubernetes liveness and
     readiness probes.
   - Logging: Using python-json-logger for JSON-formatted logs is a best practice for containerized environments,
     facilitating better observability in Azure.
   - Tooling: The greet tool is functional and correctly typed.

  2. Containerization (Dockerfile)
   - Optimization: The use of python:3.13-slim and proper layer ordering (copying requirements.txt before code)
     ensures small image sizes and fast rebuilds.
   - Environment: PYTHONUNBUFFERED=1 is correctly set to ensure logs are emitted immediately.

  3. Orchestration (k8s.yaml)
   - Probes: Well-defined liveness and readiness probes ensure high availability.
   - Resources: CPU and memory limits/requests are specified, which prevents resource exhaustion in the cluster.
   - Security: The securityContext dropping all capabilities is a strong security measure.
   - Improvement: Consider adding runAsNonRoot: true and allowPrivilegeEscalation: false to further harden the
     deployment.

  4. Automation (Makefile)
   - AKS Focus: The recent cleanup has effectively focused the automation on the AKS workflow.
   - ACR Integration: The aks-create target correctly uses --attach-acr, simplifying permission management between
     AKS and the registry.
   - Cleanup: The new destroy target provides a simple way to tear down all resources.
   - Note: The AZ_ACR_NAME uses a dynamic hostname-based suffix, which is a clever way to avoid global naming
     collisions in Azure.

  5. Dependencies (requirements.txt)
   - Recommendation: Currently, dependencies are not version-pinned. For consistent deployments, it's recommended
     to pin versions (e.g., fastmcp==0.4.1).

  6. Project Documentation (GEMINI.md & README.md)
   - Consistency: Both files are now aligned with the AKS-only deployment strategy.

  ---

  Final Verdict: The project is in excellent shape for an AKS-based MCP server. The transition from ACA to AKS is
  complete and well-documented.

Summary

The strategy for using Python for MCP development with Gemini CLI was validated with a incremental step by step approach.

A minimal HTTP transport MCP Server was started from Python source code and validated with Gemini CLI running as a MCP client in the same local environment. Then this solution was deployed remotely to the Azure Kubernates Service and validated with the local installation of Gemini CLI.

This approach can be extended to more complex deployments using other MCP transports and Cloud based options.

Deploying ADK Agents on Azure ACA (Azure Container Apps)

xbill — Thu, 02 Apr 2026 18:18:19 +0000

Leveraging the Google Agent Development Kit (ADK) and the underlying Gemini LLM to build Agents in the Python programming language deployed on the Azure Cloud platform.

What is Gemini CLI?

The full details on Gemini CLI are available here:

Build, debug & deploy with AI

Azure Container App Service

https://azure.microsoft.com/en-us/products/container-apps

Why would I want Gemini CLI with Azure? Isn’t that a Google Thing?

Azure Container App Configuration

To configure your Azure Service with the base system tools- this article provides a reference:

MCP Development with Python, and Azure Container Apps

Python Version Management

One of the downsides of the wide deployment of Python has been managing the language versions across platforms and maintaining a supported version.

The pyenv tool enables deploying consistent versions of Python:

GitHub - pyenv/pyenv: Simple Python version management

As of writing — the mainstream python version is 3.13. To validate your current Python:

xbill@penguin:~/gemini-cli-azure/adkui-aca$ python --version
Python 3.13.12

Gemini CLI

If not pre-installed you can download the Gemini CLI to interact with the source files and provide real-time assistance:

npm install -g @google/gemini-cli

Testing the Gemini CLI Environment

Once you have all the tools and the correct Node.js version in place- you can test the startup of Gemini CLI. You will need to authenticate with a Key or your Google Account:

azureuser@azure-new:~/gemini-cli-azure$ gemini

▝▜▄ Gemini CLI v0.33.1
    ▝▜▄
   ▗▟▀ Logged in with Google /auth
  ▝▀ Gemini Code Assist Standard /upgrade

Node Version Management

Gemini CLI needs a consistent, up to date version of Node. The nvm command can be used to get a standard Node environment:

GitHub - nvm-sh/nvm: Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions

To verify the Node version:

xbill@penguin:~/gemini-cli-azure/adkui-aca$ nvm --version
0.40.3
xbill@penguin:~/gemini-cli-azure/adkui-aca$ node --version
v25.9.0
xbill@penguin:~/gemini-cli-azure/adkui-aca$ npm --version
11.12.1
xbill@penguin:~/gemini-cli-azure/adkui-aca$

Agent Development Kit

The ADK can be installed from here:

Agent Development Kit (ADK)

Where do I start?

The strategy for starting Agent development is a incremental step by step approach.

The agents in the demo are based on the original code lab:

Create and deploy low code ADK (Agent Deployment Kit) agents using ADK Visual Builder | Google Codelabs

First, the basic development environment is setup with the required system variables, and a working Gemini CLI configuration.

Then, a minimal ADK Agent is built with the visual builder. Next — the entire solution is deployed to Google Cloud Run.

Setup the Basic Environment

At this point you should have a working Python environment and a working Gemini CLI installation. The next step is to clone the GitHub samples repository with support scripts:

cd ~
git clone https://github.com/xbill9/gemini-cli-azure

Then run init.sh from the cloned directory.

The script will attempt to determine your shell environment and set the correct variables:

source init.sh

If your session times out or you need to re-authenticate- you can run the set_env.sh script to reset your environment variables:

source set_env.sh

Variables like PROJECT_ID need to be setup for use in the various build scripts- so the set_env script can be used to reset the environment if you time-out.

Azure Login

All of the deployment is done with the az CLI tool- but it also needs an active login on the Azure environment. To check the version:

xbill@penguin:~/gemini-cli-azure/adkui-aca$ az --version
azure-cli 2.84.0

core 2.84.0
telemetry 1.1.0

Dependencies:
msal 1.35.0b1
azure-mgmt-resource 24.0.0

Python location '/opt/az/bin/python3'
Config directory '/home/xbill/.azure'
Extensions directory '/home/xbill/.azure/cliextensions'

Python (Linux) 3.13.11 (main, Feb 25 2026, 02:29:30) [GCC 12.2.0]

Legal docs and information: aka.ms/AzureCliLegal

Your CLI is up-to-date.
xbill@penguin:~/gemini-cli-azure/adkui-aca$

To login to Azure:

az login

Then you will see a message after successful authentication:

You have logged into Microsoft Azure!
You can close this window, or we will redirect you to the Azure CLI documentation in 1 minute.

Verify The ADK Installation

To verify the setup, run the ADK CLI with Agent3:

xbill@penguin:~/gemini-cli-azure/adkui-aca$ adk run Agent3
Log setup complete: /tmp/agents_log/agent.20260402_131951.log
To access latest log: tail -F /tmp/agents_log/agent.latest.log
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/utils/agent_loader.py:277: UserWarning: [EXPERIMENTAL] _load_from_yaml_config: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  if root_agent := self._load_from_yaml_config(actual_agent_name, agents_dir):
/home/xbill/.local/lib/python3.13/site-packages/google/adk/features/_feature_decorator.py:81: UserWarning: [EXPERIMENTAL] feature FeatureName.AGENT_CONFIG is enabled.
  check_feature_enabled()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/cli.py:204: UserWarning: [EXPERIMENTAL] InMemoryCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  credential_service = InMemoryCredentialService()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/auth/credential_service/in_memory_credential_service.py:33: UserWarning: [EXPERIMENTAL] BaseCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  super(). __init__ ()
Running agent studio_director, type exit to exit.

[studio_director]: Hello! I'm the Studio Director, ready to help you create a comic book. What kind of comic book would you like to create?




  
  
  Running the ADK Web Interface


First- start the ADK interface:



xbill@penguin:~/gemini-cli-azure/adkui-aca$ adk web --host 0.0.0.0
2026-04-02 13:21:40,287 - INFO - service_factory.py:266 - Using in-memory memory service
2026-04-02 13:21:40,287 - INFO - local_storage.py:84 - Using per-agent session storage rooted at /home/xbill/gemini-cli-azure/adkui-aca
2026-04-02 13:21:40,287 - INFO - local_storage.py:110 - Using file artifact service at /home/xbill/gemini-cli-azure/adkui-aca/.adk/artifacts
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/fast_api.py:192: UserWarning: [EXPERIMENTAL] InMemoryCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  credential_service = InMemoryCredentialService()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/auth/credential_service/in_memory_credential_service.py:33: UserWarning: [EXPERIMENTAL] BaseCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  super(). __init__ ()
INFO: Started server process [11290]
INFO: Waiting for application startup.







  
  
  Connecting to the ADK


The ADK web interface is started with the place holder 0.0.0.0 address — which tells the service to listen on all interfaces. Depending on your local environment setup- you may need to use the actual local interface. There are 2 potential options:




http://0.0.0.0:8000

http://127.0.0.1:8000






This will bring up the ADK UI:




  
  
  Run a Sub Agent


Connect to the ADK and select “Agent3”:



The first interaction just verifies the connection to the agent:




what do you do?

I manage a linear pipeline of four ADK Sequential Agents to transform a user's seed idea into a fully rendered, responsive HTML5 comic book. If your request starts with "Create me a comic of...", I'll delegate the task to my sub-agent to begin the comic creation pipeline. Otherwise, I'll answer your question directly.






Give the Agent this prompt:




Create a comic of Bob and Clippy in a Battle Royal with Tux.






The sub agent will then generate the comic book images:




  
  
  Visual Build an Agent


To use the ADK visual builder- select the pencil Icon next to Agent 2. You can drill down into the Agent design:




  
  
  What did I just do?


The initial steps validate that the Python code and the ADK environment is functional and that the Agent code will run correctly. The next step is to deploy this environment to the Azure Container App Service.


  
  
  Deploying to Azure Container App Service


To start the deployment:




xbill@penguin:~/gemini-cli-azure/adkui-aca$ make deploy
./deploy-aca.sh
=== Azure Container Apps Deployment ===
Resource Group: adk-rg-westus2
Location: westus2
ACR Name: adkacrpenguinv2
App Name: adk-app-penguin
Env Name: adk-env-penguin
Image: adkacrpenguinv2.azurecr.io/adk-image:latest
========================================






You can validate the final result by checking the messages:




│ Deploying to Azure Container App... │
│ Adding registry password as a secret with name "adkacrpenguinv2azurecrio-adkacrpenguinv2" │
│ │
│ Container app created. Access your app at │
│ https://adk-app-penguin.delightfulbeach-bf25c97b.westus2.azurecontainerapps.io/   






You can then get the endpoint:





xbill@penguin:~/gemini-cli-azure/adkui-aca$ make endpoint
--- Azure Container App Endpoint ---
adk-app-penguin.delightfulbeach-bf25c97b.westus2.azurecontainerapps.io






You can verify the service in the Azure console:




  
  
  Running the ADK Web Interface


Start a connection to the Azure Deployed ADK:




 https://adk-app-penguin.delightfulbeach-bf25c97b.westus2.azurecontainerapps.io






This will bring up the ADK UI:



Select the sub-agent “Agent3” and ask the Agent to create a comic about Bit:




Bit the Raccoon is a popular, unofficial community mascot for Microsoft Azure,
specifically representing the Cloud Advocate team and open-source initiatives. 
Created by Ashley McNamara, Bit is often depicted in tech-related, helpful 
scenarios to make complex cloud concepts more accessible and approachable 
to developers. X X +2 Role: Known as the "Developer Advocate Mascot," 
Bit is often used in presentations and content to represent the "open" 
aspect of Azure. Context: While Azure has an official anime-style mascot 
in Japan (Claudia Madobe), Bit is beloved within the broader developer 
community as a fun, tech-savvy character.

Appearance: Bit is usually styled as an adorable, helpful raccoon, 
sometimes accompanied by other community mascots like "Brady the Ninja Sloth."
GitHub GitHub Bit is a favorite for stickers and frequently appears 
on social media to highlight new Azure features.

Create a story about bit






The Agent will get to work:




  
  
  Run the Viewer Agent


Connect to the ADK and select “Agent4”:




I've exported the comic as ADK artifacts for you!

You can now view the comic directly within the ADK environment. Please check the 'Artifacts' pane or tab in your ADK Visual Builder / Studio UI.

You'll find the following:

view_full_comic.html: This is a complete, self-contained HTML file with embedded images for the full comic experience.
comic_preview.md: A markdown summary of the comic.
Individual panel images (e.g., panel_panel_1.png).
If you are working in a local terminal, you can still view the original local files by opening output/comic.html in your browser or by running ./comic.sh. However, the ADK artifacts are the best way to view it while hosted!






The Agent will generate the artifacts:



The Comic can then be viewed:




  
  
  Summary


The Agent Development Kit was used to visually define a basic agent and added a tool that calls the ImageGen Gemini Model. This Agent was tested locally with the CLI and then with the ADK web tool. Then, the Agent was packaged and deployed to the Azure Container App Service. Several sample ADK agents were run directly from the deployment in Azure. This approach validates that cross cloud tools can be used — even with more complex agents.

Deploying ADK Agents on Azure ACI

xbill — Thu, 02 Apr 2026 17:00:19 +0000

Leveraging the Google Agent Development Kit (ADK) and the underlying Gemini LLM to build Agents in the Python programming language deployed securely on the Azure Cloud platform.

What is Gemini CLI?

The full details on Gemini CLI are available here:

Build, debug & deploy with AI

Azure Container Instances

More details are available here:

https://azure.microsoft.com/en-us/products/container-instances

Why would I want Gemini CLI with Azure? Isn’t that a Google Thing?

Azure Container Instance Configuration

To configure your Azure Container Instances with the base system tools- this article provides a reference:

MCP Development with Python, and the Azure Container Instance

Python Version Management

One of the downsides of the wide deployment of Python has been managing the language versions across platforms and maintaining a supported version.

The pyenv tool enables deploying consistent versions of Python:

GitHub - pyenv/pyenv: Simple Python version management

As of writing — the mainstream python version is 3.13. To validate your current Python:

xbill@penguin:~/gemini-cli-azure/adkui-appservice$ python --version
Python 3.13.12

Gemini CLI

If not pre-installed you can download the Gemini CLI to interact with the source files and provide real-time assistance:

npm install -g @google/gemini-cli

Testing the Gemini CLI Environment

Once you have all the tools and the correct Node.js version in place- you can test the startup of Gemini CLI. You will need to authenticate with a Key or your Google Account:

azureuser@azure-new:~/gemini-cli-azure$ gemini

▝▜▄ Gemini CLI v0.33.1
    ▝▜▄
   ▗▟▀ Logged in with Google /auth
  ▝▀ Gemini Code Assist Standard /upgrade

Node Version Management

Gemini CLI needs a consistent, up to date version of Node. The nvm command can be used to get a standard Node environment:

GitHub - nvm-sh/nvm: Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions

To verify the Node version:

xbill@penguin:~/gemini-cli-azure/adkui-aci$ node --version
v25.9.0
xbill@penguin:~/gemini-cli-azure/adkui-aci$ nvm --version
0.40.3
xbill@penguin:~/gemini-cli-azure/adkui-aci$ npm --version
11.12.1

Agent Development Kit

The ADK can be installed from here:

Agent Development Kit (ADK)

Where do I start?

The strategy for starting Agent development is a incremental step by step approach.

The agents in the demo are based on the original code lab:

Create and deploy low code ADK (Agent Deployment Kit) agents using ADK Visual Builder | Google Codelabs

First, the basic development environment is setup with the required system variables, and a working Gemini CLI configuration.

Then, a minimal ADK Agent is built with the visual builder. Next — the entire solution is deployed to Google Cloud Run.

Setup the Basic Environment

At this point you should have a working Python environment and a working Gemini CLI installation. The next step is to clone the GitHub samples repository with support scripts:

cd ~
git clone https://github.com/xbill9/gemini-cli-azure

Then run init.sh from the cloned directory.

The script will attempt to determine your shell environment and set the correct variables:

source init.sh

If your session times out or you need to re-authenticate- you can run the set_env.sh script to reset your environment variables:

source set_env.sh

Variables like PROJECT_ID need to be setup for use in the various build scripts- so the set_env script can be used to reset the environment if you time-out.

Azure Login

All of the deployment is done with the az CLI tool- but it also needs an active login on the Azure environment. To check the version:

xbill@penguin:~/gemini-cli-azure/adkui-aci$ az --version
azure-cli 2.84.0

core 2.84.0
telemetry 1.1.0

Dependencies:
msal 1.35.0b1
azure-mgmt-resource 24.0.0

Python location '/opt/az/bin/python3'
Config directory '/home/xbill/.azure'
Extensions directory '/home/xbill/.azure/cliextensions'

Python (Linux) 3.13.11 (main, Feb 25 2026, 02:29:30) [GCC 12.2.0]

Legal docs and information: aka.ms/AzureCliLegal

Your CLI is up-to-date.

To login to Azure:

az login
You have logged into Microsoft Azure!
You can close this window, or we will redirect you to the Azure CLI documentation in 1 minute.

Verify The ADK Installation

To verify the setup, run the ADK CLI with Agent3:

xbill@penguin:~/gemini-cli-azure/adkui-aci$ adk run Agent3
Log setup complete: /tmp/agents_log/agent.20260402_120259.log
To access latest log: tail -F /tmp/agents_log/agent.latest.log
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/utils/agent_loader.py:277: UserWarning: [EXPERIMENTAL] _load_from_yaml_config: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  if root_agent := self._load_from_yaml_config(actual_agent_name, agents_dir):
/home/xbill/.local/lib/python3.13/site-packages/google/adk/features/_feature_decorator.py:81: UserWarning: [EXPERIMENTAL] feature FeatureName.AGENT_CONFIG is enabled.
  check_feature_enabled()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/cli.py:204: UserWarning: [EXPERIMENTAL] InMemoryCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  credential_service = InMemoryCredentialService()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/auth/credential_service/in_memory_credential_service.py:33: UserWarning: [EXPERIMENTAL] BaseCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  super(). __init__ ()
Running agent studio_director, type exit to exit.



  
  
  Running the ADK Web Interface


First- start the ADK interface:



xbill@penguin:~/gemini-cli-azure/adkui-aci$ adk web --host 0.0.0.0
2026-04-02 12:03:34,525 - INFO - service_factory.py:266 - Using in-memory memory service
2026-04-02 12:03:34,525 - INFO - local_storage.py:84 - Using per-agent session storage rooted at /home/xbill/gemini-cli-azure/adkui-aci
2026-04-02 12:03:34,525 - INFO - local_storage.py:110 - Using file artifact service at /home/xbill/gemini-cli-azure/adkui-aci/.adk/artifacts
/home/xbill/.local/lib/python3.13/site-packages/google/adk/cli/fast_api.py:192: UserWarning: [EXPERIMENTAL] InMemoryCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  credential_service = InMemoryCredentialService()
/home/xbill/.local/lib/python3.13/site-packages/google/adk/auth/credential_service/in_memory_credential_service.py:33: UserWarning: [EXPERIMENTAL] BaseCredentialService: This feature is experimental and may change or be removed in future versions without notice. It may introduce breaking changes at any time.
  super(). __init__ ()
INFO: Started server process [5030]
INFO: Waiting for application startup.

+-----------------------------------------------------------------------------+
| ADK Web Server started |
| |
| For local testing, access at http://0.0.0.0:8000. |
+-----------------------------------------------------------------------------+

INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)







  
  
  Connecting to the ADK


The ADK web interface is started with the place holder 0.0.0.0 address — which tells the service to listen on all interfaces. Depending on your local environment setup- you may need to use the actual local interface. There are 2 potential options:




http://0.0.0.0:8000 

http://127.0.0.1:8000






This will bring up the ADK UI:




  
  
  Run a Sub Agent


Connect to the ADK and select “Agent3”:



The first interaction just verifies the connection to the agent:




what do you do?

I manage a linear pipeline of four ADK Sequential Agents to transform a user's seed idea into a fully rendered, responsive HTML5 comic book. If your request starts with "Create me a comic of...", I'll delegate the task to my sub-agent to begin the comic creation pipeline. Otherwise, I'll answer your question directly.






Give the Agent this prompt:




Create a comic of Bob and Clippy in a Battle Royal with Tux.






The sub agent will then generate the comic book images:




  
  
  Visual Build an Agent


To use the ADK visual builder- select the pencil Icon next to Agent 2. You can drill down into the Agent design:




  
  
  What did I just do?


The initial steps validate that the Python code and the ADK environment is functional and that the Agent code will run correctly. The next step is to deploy this environment to the Azure Container Instance Service.


  
  
  Deploying to Azure Container Instance


To start the deployment:




xbill@penguin:~/gemini-cli-azure/adkui-aci$ make deploy
Building the Docker image...
docker build -t adk-image .






You can validate the final result by checking the messages:




✦ The deployment to Azure Container Instance (ACI) was successful. 

  The application is now accessible at: http://adk-container-penguin.westus2.azurecontainer.io:8080






You can then get the endpoint:




xbill@penguin:~/gemini-cli-azure/adkui-aci$ make endpoint
--- Azure Container Instance Endpoint ---
adk-container-penguin.westus2.azurecontainer.io
xbill@penguin:~/gemini-cli-azure/adkui-aci$ 






The service will be visible in the Azure console:




  
  
  Running the ADK Web Interface


Start a connection to the Azure Deployed ADK:



This will bring up the ADK UI.

Select the sub-agent “Agent3” and ask the Agent to create a comic about Bit:




Bit the Raccoon is a popular, unofficial community mascot for Microsoft Azure,
specifically representing the Cloud Advocate team and open-source initiatives. 
Created by Ashley McNamara, Bit is often depicted in tech-related, helpful 
scenarios to make complex cloud concepts more accessible and approachable 
to developers. X X +2 Role: Known as the "Developer Advocate Mascot," 
Bit is often used in presentations and content to represent the "open" 
aspect of Azure. Context: While Azure has an official anime-style mascot 
in Japan (Claudia Madobe), Bit is beloved within the broader developer 
community as a fun, tech-savvy character.  

Appearance: Bit is usually styled as an adorable, helpful raccoon, 
sometimes accompanied by other community mascots like "Brady the Ninja Sloth."
GitHub GitHub Bit is a favorite for stickers and frequently appears 
on social media to highlight new Azure features.  






The Agent will get to work:




  
  
  Run the Viewer Agent


Connect to the ADK and select “Agent4”:




I've exported the comic as ADK artifacts for you!

You can now view the comic directly within the ADK environment. Please check the 'Artifacts' pane or tab in your ADK Visual Builder / Studio UI.

You'll find the following:

view_full_comic.html: This is a complete, self-contained HTML file with embedded images for the full comic experience.
comic_preview.md: A markdown summary of the comic.
Individual panel images (e.g., panel_panel_1.png).
If you are working in a local terminal, you can still view the original local files by opening output/comic.html in your browser or by running ./comic.sh. However, the ADK artifacts are the best way to view it while hosted!






The Agent will generate the artifacts:



The Comic can then be viewed:




  
  
  Summary


The Agent Development Kit was used to visually define a basic agent and added a tool that calls the ImageGen Gemini Model. This Agent was tested locally with the CLI and then with the ADK web tool. Then, the Agent was packaged and deployed to the Azure Container Instance. Several sample ADK agents were run directly from the deployment in Azure. This approach validates that cross cloud tools can be used — even with more complex agents.