AI & Agent Workflows

Overview

AI-assisted development workflows need a fast and secure environment to develop and test in. Whether you’re using an AI coding assistant to generate infrastructure code, running an agent that deploys AWS resources, or validating AI-generated IaC before applying it to real AWS — LocalStack gives you a safe, fast, cost-free environment to run these workflows in.

Connect an AI coding assistant via MCP

The LocalStack MCP Server is a Model Context Protocol (MCP) server that connects compatible clients (Claude, Cursor, Windsurf, and others) to your LocalStack environment. It gives agents tools to run the full local cloud lifecycle—not only generating infrastructure code, but starting the container, deploying and tearing down IaC, analyzing logs, and more. See Introducing the LocalStack MCP Server for the original announcement and walkthrough.

With the MCP server configured, your assistant can use tools such as:

• Lifecycle — Start, stop, restart, or check status of the LocalStack container
• Deploy — Deploy or destroy CDK, Terraform, SAM, or CloudFormation projects on LocalStack
• Operations — Analyze logs, run AWS CLI commands via awslocal, manage Cloud Pods, and (where supported) IAM policy analysis, chaos injection, and docs search

…and many more.

MCP Server Prerequisites

Before you begin, make sure these are installed and in your system’s PATH:

• Node.js (v20.x or later) to run the npx command.
• LocalStack CLI and Docker to manage the LocalStack container itself.
• cdklocal or tflocal to use the localstack-deployer tool. (optional)
• A LocalStack Auth Token to enable licensed features. (optional)

MCP Server Configuration

Add a server entry like this to your client’s MCP configuration—for Cursor, this would typically be ~/.cursor/mcp.json. For Claude Code, register the server with claude mcp add (it is saved to your project-level MCP configuration).

{
  "mcpServers": {
    "localstack-mcp-server": {
      "command": "npx",
      "args": ["-y", "@localstack/localstack-mcp-server"],
      "env": {
        "LOCALSTACK_AUTH_TOKEN": "<YOUR_TOKEN>"
      }
    }
  }
}

Full prerequisites, Cursor/VS Code/Claude Code variants, custom endpoints (LOCALSTACK_HOSTNAME, LOCALSTACK_PORT), troubleshooting, and the complete tool reference are on LocalStack MCP Server.

Note

You do not need to start LocalStack manually before using MCP: an agent can start the container using the localstack-management tool once this configuration is in place. If LocalStack already runs elsewhere, point the MCP server at it via the env block as described in Connecting to a custom LocalStack endpoint.

Deploy with agent-driven automation using Skills

LocalStack Skills are pre-built agent skill definitions for deploying common AWS architectures locally. It contains skills for LocalStack lifecycle management, IaC deployment, state management and more. Skills can be used in combination with the MCP server to create a fully autonomous agent-driven workflow.

Skills are useful when:

• You want to scaffold a new local environment quickly without writing all the infrastructure code yourself
• You’re using an agent-first workflow and want LocalStack to be a first-class deployment target
• You want to iterate rapidly on architecture without touching real AWS

Browse the skills repository for available skills and setup instructions.

Validate AI-generated IaC before applying to AWS

A common pattern when using AI to generate Terraform, CDK, or CloudFormation is to deploy it to LocalStack first. This catches configuration errors, missing permissions, and service interaction bugs before you spend time (and money) deploying to real AWS.

The workflow is:

1. Generate infrastructure code with your AI tool
2. Deploy to LocalStack with tflocal apply (Terraform), cdklocal deploy (CDK), or the AWS CLI
3. Run your integration tests against the local environment
4. When everything passes, deploy to real AWS with confidence

See Tooling for the full list of LocalStack-aware wrappers for common IaC tools.

Summary

Use case	Tool
AI assistant that can inspect & manage local resources	LocalStack MCP Server
Agent-driven infrastructure deployment	LocalStack Skills
Validate AI-generated IaC safely	LocalStack + tflocal / cdklocal / awslocal