// Stay in touch?
Products
Encore CloudEncore Cloud
Encore.tsEncore.ts
Encore.goEncore.go
PricingPricing
Book a DemoBook a Demo
Use Cases
AI-Powered DevelopmentAI-Powered Development
Event-Driven SystemsEvent-Driven Systems
Distributed SystemsDistributed Systems
Case StudiesCase Studies
ShowcaseShowcase
Resources
DocsDocs
InstallInstall
Example AppsExample Apps
Demo videoDemo video
ArticlesArticles
ResourcesResources
GitHub ReleasesGitHub Releases
Systems Operational
Company
About UsAbout Us
Swag ShopSwag Shop
ContactContact
JobsJobs
PressPress
TermsTerms
Privacy PolicyPrivacy Policy
Data Processing AgreementData Processing Agreement
Enterprise SLAEnterprise SLA
Encore
© 2026 EncoreAll rights reserved
© 2026 Encore All Rights Reserved
GitHubDiscordYouTube

Leap is here: AI dev agent that builds Encore.ts apps and deploys to your own cloud

Combining AI with Encore.ts for incredible productivity

May 31, 2025
3 Min Read
Marcus Kohlberg
May 31, 2025

Leap is here: AI dev agent that builds Encore.ts apps and deploys to your own cloud

Combining AI with Encore.ts for incredible productivity

Marcus Kohlberg
3 Min Read

This was originally posted on the Leap website on May 28, 2025.

We're excited to announce Leap, an AI developer agent that helps you build and ship production-grade apps to your own cloud.

We started building Leap after trying every other AI app builder and hitting the same limitations: they only supported single-page apps, locked you into proprietary Backend-as-a-Service platforms, and lacked support for real production workflows.

Most of these tools don't even offer proper isolated test environments, making deployment risky. Approving a buggy SQL migration, for example, could affect your live database.

That might be fine for some use cases. But as developers, we needed support for advanced infrastructure, more complex architectures, and a more robust workflow.

To be fair, these limitations aren't necessarily flaws. Many of these tools are built for non-developers who value simplicity over flexibility.

We wanted something different.

We wanted a tool for developers, something that delivers the speed of AI without sacrificing control over architecture, APIs, infrastructure, or deployment.

That's why we built Leap.

What makes Leap different

You still start from a prompt. But everything after that is developer-first:

  • You iterate using changes and revisions. Each change is like a pull request, grouping together a set of related revisions, while each revision is like a commit, capturing the specific modifications made. Every revision produces a diff so you can review and approve exactly what's changed. (Manual code edits are, of course, fully supported.)
  • A GitHub integration lets you own and control your codebase from the start.
  • Leap keeps you in control with live architecture diagrams and structured API documentation, helping you understand what the AI is building and making it easy to share that knowledge with others.
  • Leap's preview environment is completely isolated from your other environments, so you can test with confidence.
  • For deployment, you can use Encore.ts's open-source tools to package your app as a Docker container and deploy anywhere. Or use the integration with Encore Cloud to automate deployments and infrastructure provisioning, directly into your AWS or GCP cloud.

How Leap works

  • Leap uses Claude 4 Sonnet for most of the code generation, and also uses some other models like Gemini 2.5 for parts of the process.
  • Apps built with Leap use Encore.ts for the backend. It's our open-source framework, developed over the past few years, trusted by thousands of developers and with over 11k GitHub stars. The framework includes a declarative infrastructure layer, like a cloud-agnostic version of AWS CDK, which Leap uses to set up infrastructure for microservices, databases, pub/sub, storage buckets, and more.
  • Leap's isolated preview environments are set up using Firecracker running on Hetzner.

Where Leap fits in

Leap isn't meant to replace every other workflow or tool right away. We expect it to be most useful for spinning up new projects or building isolated systems within an existing product. You can easily move over to your regular IDE whenever you feel like it, using the GitHub integration to control your codebase.

We built Leap because nothing else met our needs as developers. Now that we're live and starting to welcome users, we'd love to hear your thoughts.

  • Does this solve any development or deployment pain you've experienced?
  • What would you need to confidently use this to build production-grade apps?

Your feedback will help shape where Leap goes next. We hope you'll give it a try and let us know what you think.

Try Leap

Leap comes with limited free usage and is available for you to try right now.
Start Building with Leap →

We can't wait to see what you build!

– The Encore / Leap team

Questions? Join us on Discord or email us at [email protected].

Contents
What makes Leap different
How Leap works
Where Leap fits in
Try Leap

A development platform for your own cloud on AWS & GCP

Encore automates infrastructure management, observability, and documentation. Your team can focus on shipping product.

Encore

This blog is presented by Encore, the backend framework for building robust type-safe distributed systems with declarative infrastructure.

Like this article?
Get future ones straight to your mailbox.

You can unsubscribe at any time.

Related Articles

AI
06/03/26 / 7 Min Read
AI
06/03/26 / 7 Min Read
AI agents love type errors
An agent only fixes what it can see before it stops, and a compile error is the one signal that always lands in time.
Ivan Cernja
AI
05/20/26 / 14 Min Read
AI
05/20/26 / 14 Min Read
Are TypeScript backend frameworks ready for AI agents?
We set out to run one benchmark across five TypeScript backend frameworks. Reading the diffs sent us into two more runs, and the picture changed each time.
Ivan Cernja
Infrastructure
05/04/26 / 7 Min Read
Infrastructure
05/04/26 / 7 Min Read
What is distributed tracing?
An interactive guide to debugging real backend issues
Ivan Cernja