Introduction

What is Agent Cloud?

Agent Cloud (cloud-agent) is an open-source, AI-powered command-line tool that takes any software project and deploys it to AWS, Google Cloud, or Azure. Instead of manually writing Dockerfiles, configuring cloud services, and running dozens of CLI commands, you run a single command and let AI agents handle the complexity.

The tool is built in TypeScript using the Mastra AI framework and uses three specialized agents that work together:

1. Analyzer Agent — scans your codebase to detect the technology stack, framework, databases, and project type
2. Deployment Planner Agent — maps your project to the optimal cloud services, estimates costs, and generates deployment commands
3. Validator Agent — checks that your local environment is properly configured (CLI tools installed, authenticated, correct permissions)

Why Agent Cloud?

Deploying a project to the cloud typically involves:

• Figuring out which cloud services fit your project (ECS vs. Lambda vs. App Service vs. Cloud Run)
• Writing a Dockerfile or configuring a build pipeline
• Setting up IAM roles, security groups, and networking
• Running a series of CLI commands in the right order
• Debugging cryptic error messages when something goes wrong

Agent Cloud collapses all of that into a guided workflow. The AI reads your code the same way an experienced DevOps engineer would, then generates a complete deployment plan. You review and approve the plan before anything is executed.

How It Works

The deployment flow has five phases:

1. Environment Validation
   The validator agent checks your cloud CLI, authentication, and permissions.

2. Project Analysis
   The analyzer agent scans your project directory to detect runtime,
   framework, dependencies, databases, and Docker configuration.

3. Deployment Planning
   The deployment agent maps your project to cloud services, estimates
   monthly costs, and generates the exact CLI commands.

4. Human Approval (suspend point)
   The workflow pauses and presents the plan. You see the services,
   estimated cost, and commands. You approve or reject.

5. Cloud Execution
   If approved, the tool runs the actual cloud CLI commands to create
   resources and deploy your application.

Core Principles

Local-first analysis. The analyze command works without any API key. It reads package.json, requirements.txt, go.mod, and other manifest files directly. AI-powered deep analysis is opt-in with the --ai flag.

Human-in-the-loop. No cloud resources are created without your explicit approval. The workflow uses Mastra’s suspend/resume mechanism to pause before execution.

Multi-cloud by design. Every analysis produces recommendations for all three major cloud providers, so you can compare services and costs before choosing.

Real deployments. This is not a dry-run tool. When you approve a plan, it executes real cloud CLI commands (aws, gcloud, az) to create infrastructure and deploy your application.

Supported Project Types

Agent Cloud can detect and deploy the following project types:

Type	Examples	AWS	GCP	Azure
API	Express, FastAPI, NestJS, Hono	ECS Fargate, Lambda	Cloud Run, Cloud Functions	Container Apps, Functions
Web App	Next.js, React, Vue, Svelte	ECS Fargate, Amplify	Cloud Run, App Engine	App Service, Static Web Apps
Static Site	HTML/CSS/JS, Astro, Gatsby	S3 + CloudFront	Firebase Hosting	Static Web Apps
Container	Any Dockerfile	ECS Fargate	Cloud Run	Container Apps

Supported Languages

• Node.js — npm, yarn, pnpm, and bun package managers
• Python — pip with requirements.txt
• Go — go.mod detection
• Any language with a Dockerfile

Next Steps

• Install Agent Cloud to get the CLI on your machine
• Quick Start to deploy your first project in three commands
• Configuration to set up API keys and preferences