Gemini CLI: Write and deploy a Cloud Run app in 5 minutes

Key Concepts

• Gemini CLI: An open-source AI agent that provides direct access to Gemini models within the terminal.
• MCP Server (Model-Centric Programming Server): A feature that allows plain English commands to interact with APIs, simplifying complex operations.
• Cloud Run: A fully managed serverless platform on Google Cloud that enables developers to run stateless containers.
• Cloud Workstations: A managed development environment in the cloud.
• Python Flask: A lightweight Python web framework.
• Dockerfile: A text document that contains all the commands a user could call on the command line to assemble an image.
• Gunicorn: A Python WSGI HTTP Server for UNIX.
• YOLO Mode: A feature in Gemini CLI that allows automatic changes without user confirmation.
• Virtual Python Environment: An isolated Python environment to manage dependencies for different projects.

Introduction to Prototyping Web Apps

The video demonstrates a streamlined process for prototyping and deploying web applications, emphasizing ease of use and minimal fuss. The core idea is to leverage AI assistance directly within the developer's workflow.

Using Gemini CLI for App Development

The demonstration focuses on using the Gemini CLI, an AI agent accessible via the terminal, to build and deploy an application. The process involves:

1. Setting up Gemini CLI:

   • The user installs Gemini CLI on their machine.
   • A .gemini directory is used to store settings.
   • MCP Server Configuration: Cloud Run is added as an MCP server in the settings. This enables Gemini CLI to understand and execute Cloud Run deployment commands using plain English.

2. Code Generation with Gemini CLI:

   • A new directory (hello app) is created for the application.
   • The Gemini CLI is initiated within this directory.
   • Prompt: The user instructs Gemini CLI to "create a Python Flask app with a text box for the user's name. It should echo back 'hello' plus the person's name."
   • Gemini's Reasoning (Gemini 2.5 Pro): Gemini breaks down the request into a list of actionable steps (six in this instance) and presents them for user confirmation. This highlights Gemini's reasoning capabilities.
   • User Confirmation: The user approves each step. The option of "YOLO mode" is mentioned, which would automate these changes without explicit approval, suitable for trusted workflows.

3. Local Testing and Iteration:

   • Gemini CLI generates the necessary application files.
   • The user asks Gemini CLI to run the commands to start the app.
   • Error Handling and Debugging: An error occurs during execution. Gemini CLI identifies the need for a virtual Python environment, creates one, and resolves subsequent errors, demonstrating its debugging and problem-solving abilities.
   • The app is successfully started and runs locally on the cloud workstation.
   • The user tests the app by entering their name, confirming it echoes back "hello [name]".

Deploying to Cloud Run

Once the local prototype is functional, the next step is to deploy it to the cloud for wider accessibility.

1. Stopping the Local App: The user instructs Gemini CLI to stop the locally running application.

2. Initiating Cloud Deployment:

   • The user asks Gemini CLI to deploy the app to Cloud Run.
   • Dockerfile Generation: Gemini CLI recommends and automatically generates a Dockerfile, eliminating manual creation.
   • Dependency Management: Gemini CLI identifies the need for Gunicorn and adds it to the application.
   • Leveraging MCP Server: The pre-configured Cloud Run MCP server allows Gemini CLI to directly handle the deployment process.

3. Container Build and Deployment:

   • Gemini CLI builds the Docker container and deploys it to Cloud Run. This process takes approximately a minute.

4. Accessing the Deployed App:

   • A service URL is provided, which is the public address of the deployed application.
   • The user accesses the URL, enters their name, and confirms the app functions identically to the local version.

Post-Deployment Management

• Viewing Logs: Thanks to the MCP server integration, Gemini CLI can access and display application logs from Cloud Run. Gemini analyzes the logs and confirms the app is running well.

Cost Considerations

• Gemini CLI Free Tier: Gemini CLI offers a generous free allowance of 60 requests per minute and up to 1,000 requests per day, described as the largest free allowance in the industry. Enterprise licenses are available for higher usage.
• Cloud Run Free Tier: Cloud Run provides a free monthly quota including 180,000 CPU seconds, 360,000 GB seconds, and 2 million requests. Beyond this, users pay only when their service is actively processing requests, making low-traffic services often free.

Capabilities of Gemini CLI

Beyond code generation and deployment, Gemini CLI offers a wide range of functionalities:

• Writing code
• Creating unit tests
• Generating test data
• Explaining code functionality
• Analyzing error messages and debugging
• Writing documentation
• Identifying and executing shell commands

All these tasks can be performed by simply asking in plain English.

Gemini CLI vs. Web UI

While the Gemini web UI is also effective, the CLI is designed to be integrated into a developer's existing toolchain. Its presence in the terminal allows for seamless interaction with local files, chaining commands (e.g., with git diff), and maintaining workflow continuity without context switching.

Conclusion and Key Takeaways

The video concludes with two main takeaways:

1. Ease of App Development and Deployment: It is now easier than ever to write an application and deploy it to the cloud.
2. Enabling Technologies: This increased ease is attributed to the combined power of Gemini CLI and the Cloud Run MCP server.