Get started with infra.new

Learn how to use infra.new effectively

On this page

What is infra.new?

An AI-powered DevOps copilot that helps you configure and deploy apps on any cloud using:

  • Terraform for infrastructure configuration
  • Cloud CLIs (aws, gcloud, az) for deploying and debugging
  • GitHub Actions for CI/CD automation

The agent runs almost entirely in your browser and can optionally be configured to use a sandboxed runtime for running commands that require more dependencies.

Infra.new is built for configuring and using infrastructure tools - if you want to generate applications, we recommend using v0, lovable, or bolt then handoff the app code to infra.new to deploy to your cloud.

Capabilities at a glance

Documentation Access

Fetches the latest documentation to assist with configuration tasks

Terraform
GitHub Actions
Helm Charts
Ansible
Available now
Coming soon
Under consideration

How to use infra.new

Using infra.new effectively involves a three-step workflow:

  1. Starting with a clear prompt
  2. Refining the agent's implementation plan
  3. Stepping through the plan with the agent

Each step builds on the previous one to help guide the agent to a high quality result.

Start with a clear prompt

Begin by describing your infrastructure needs in detail. You can optionally upload your source code and attach pdfs/images to give the agent more context about your application. The agent will attempt to clarify any ambiguous requests, but the quality of the agent's output depends significantly on how well you describe what you need.

Good prompt:

"I need to set up a highly available web application on AWS using ECS Fargate. The app needs an RDS Postgres database, Redis for caching, and should be deployed across multiple availability zones."

Bad prompt:

"Deploy my app to the cloud."

Include details about cloud providers, regions, and any specific services or configurations you need.

Refine the agent's plan

The agent will analyze your requirements and either execute the task immediately or generate a plan with step-by-step instructions for it to follow. We recommend asking the agent to create a plan for most use cases and spending some time refining the steps before approving. The planning stage is the most important stage for ensuring high quality results.

Infrastructure Blueprint
Use your custom prompt settings to specify where to store your terraform state and any default requirements such as cloud regions, account IDs, project IDs, etc. This saves time and ensures consistency across your infrastructure.

Step through the plan

The agent will follow the blueprint and pause after each step, giving you a chance to review the changes and fix any issues detected by the validation checks. It's tempting to click through the steps quickly, but it's usually worth spending a minute to review the changes before moving onto the next step.

Once you're happy with your infrastructure changes, you can either export the code as a zip file to deploy your infrastructure locally, or sync all changes to GitHub to deploy using CI/CD. We recommend doing a test deployment before setting up automation to catch any configuration issues early in the process.

Runners are best suited for development and testing environments. We don't recommend using this feature in production environments.

Final tips

  • For best results, keep your conversations focused on a single task. We do our best to trim old context once your conversation gets long, but it's usually best to start with a fresh context window when starting a new unrelated task.
  • If the agent makes a bad change or incorrect assumption, it's often better to delete the last message and add more details rather than prompting the agent with a followup.

Need help?

Schedule a call

Have questions or want a personalized demo?

Loading calendar...