Leap.new — Full-Stack AI Dev Agent Deploying to AWS and GCP

TL;DR

TL;DR: Leap.new is an AI developer agent that builds full-stack applications with real backend infrastructure and deploys them to your own AWS or GCP cloud — you own the code, via a GitHub-connected PR workflow.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: leap.new — verified 2026-07-23
• Official launch: leap.new/blog/launch — verified 2026-07-23
• HN launch thread: news.ycombinator.com/item?id=44137177
• Framework: Encore.ts — open source TypeScript framework
• Last verified: 2026-07-23

What Is Leap.new?

Leap.new is an AI developer agent that helps you build and deploy full-stack applications with real backends to your own AWS or GCP cloud. The key differentiator from other AI app builders is that Leap targets developers who need production-grade infrastructure — not just prototypes:

• Code lives in your own GitHub repository — you own it
• Changes come through pull request-style revisions with diffs you review before merging
• Apps get an isolated preview environment for each change
• Deploy to your own AWS or GCP account (not Leap’s infrastructure)

Leap uses Claude 4 Sonnet as the model and runs on the open-source Encore.ts framework, which provides declarative infrastructure similar to a cloud-agnostic CDK layer.

Setup Workflow

Prerequisites

• A GitHub account (code is stored in your own repos)
• An AWS or GCP account (for deployment targets)
• A Leap.new account (free to start)

Step 1 — Connect Your GitHub

Sign in at leap.new and connect your GitHub account. Leap creates repos in your own GitHub organization or personal account, so you retain full control of the code.

Step 2 — Describe Your App

Use the prompt interface to describe the application you want to build. Leap will generate an initial architecture, then iterate based on your feedback through the PR workflow.

Step 3 — Review PR-Style Diffs

Each round of AI-generated changes appears as a pull request in your GitHub repo. You review the diff, request changes, or merge — just like normal code review.

Step 4 — Deploy

Leap can deploy via two paths:

• Encore Cloud integration — automated deploys to your AWS or GCP account via Encore’s infrastructure layer
• Open source tooling — deploy using the Encore.ts CLI directly if you prefer self-hosting the infrastructure layer

How Leap Compares to Other AI App Builders

Most AI coding agents and app builders either generate code you don’t own, use proprietary hosting you can’t escape, or produce prototypes that fall apart when you hit real infrastructure requirements.

Leap’s PR-first workflow means you always see exactly what changed before it runs anywhere. The Encore.ts backend framework handles declarative infrastructure — databases, queues, scheduled jobs, API gateways — as code, so the AI generates infrastructure alongside application logic rather than inventing ad-hoc workarounds.

The isolated preview environment per change is particularly useful for iterating quickly without breaking staging or production.

Practical Evaluation Checklist

• ✅ Builds full-stack apps (frontend + backend API + database)
• ✅ Code stored in your own GitHub repo
• ✅ PR-style diff review before any code merges
• ✅ Deploys to your own AWS or GCP account
• ✅ Isolated preview environment per change
• ✅ Uses Claude 4 Sonnet
• ✅ Runs on open-source Encore.ts framework
• ✅ Architecture diagrams and API docs auto-generated as you build
• ⚠️ Early-stage product (verify current capabilities on leap.new before relying on specific features)

Security Notes

• Code ownership — all generated code lives in your GitHub account, not Leap’s infrastructure
• Infrastructure access — Leap needs credentials to your cloud account for deployment; review the IAM permissions it requests before granting
• Open-source framework — Encore.ts is open source (MIT), so the backend runtime is auditable
• Self-hosting option — using the open-source Encore.ts CLI means you can run the infrastructure layer yourself without Leap’s managed platform

FAQ

Q: Do I need to use Leap’s managed platform to deploy? A: No. Leap runs on the open-source Encore.ts framework. You can deploy using the CLI directly to your own cloud account without the Leap managed layer.

Q: What cloud providers does Leap support? A: AWS and GCP are explicitly supported via Encore Cloud’s integration. Self-hosted Encore.ts deployments can target any provider that runs Docker containers.

Q: What model does Leap use? A: Claude 4 Sonnet, according to the launch announcement.

Q: How does the PR review workflow work? A: Leap generates code in a feature branch in your repo. A pull request is opened automatically. You review the diff, comment, request changes, or merge. The AI iterates based on your review comments.

Q: Is this open source? A: The backend framework (Encore.ts) is open source under the MIT license. Leap.new itself is a hosted product built on top of it.

Q: What kind of apps can I build with Leap? A: Full-stack web apps with REST or GraphQL APIs, databases, background jobs, and scheduled tasks. The Encore.ts framework covers most standard backend patterns.

Conclusion

Leap.new is worth watching if you want an AI coding agent that produces production-ready backend infrastructure rather than prototype code. The PR-first review model and GitHub-native workflow make it fit naturally into existing development teams, and the Encore.ts open-source backend gives you an exit ramp if you outgrow the managed platform.

The key question for any AI dev agent is whether the generated infrastructure code is actually correct and secure — the Encore.ts framework’s declarative model helps here by constraining what the AI can generate. Worth trying on a small real project to evaluate.