There's a moment most engineering teams recognize. You need to ship a backend service. Something straightforward, even—an API, a few database tables, authentication. Three hours later, you're wrist-deep in Terraform configurations, toggling between AWS consoles, configuring CI/CD pipelines, and wondering why it takes six separate tools just to get to the starting line. You still haven't written actual application logic.
André Eriksson and Marcus Kohlberg, who cut their teeth at Spotify, think that workflow is fundamentally broken. Their company, Encore, has spent the past three years building what they describe as the antidote: a platform that lets infrastructure declarations live directly inside application code, eliminating the usual sprawl of deployment tools. In late May, they added an AI layer called Leap—not a code autocomplete tool, but an agent designed to generate production-grade backends, infrastructure definitions included.
The pitch: one platform, your own cloud account, no Terraform required.
Whether that vision holds up under scrutiny depends on whether consolidation can deliver what modularity promised and frequently failed to provide—speed without sacrificing control.
Infrastructure That Reads Your Code
Encore Cloud operates on a premise that sounds simple until you consider the plumbing involved. Developers write TypeScript or Go (Python support is in the works), defining their infrastructure needs directly within application code. The platform parses that code, figures out what you need—databases, pub/sub queues, object storage, secrets—and provisions everything in your own AWS or GCP account.
No separate Terraform files. No Helm charts living in a different repository. No context-switching between your IDE and seven browser tabs.
The company raised $3 million in seed funding back in April 2022, led by Crane Venture Partners, which suggests early investors saw potential in collapsing what has become an increasingly bloated development toolchain.
What arrives out of the box is more extensive than you might expect. Distributed tracing, metrics, API explorers, architecture diagrams that update themselves, service catalogs—all generated automatically. For GCP deployments, Encore wires up Cloud Run or Compute Engine instances, Cloud SQL databases, Cloud Storage, plus the observability stack through Cloud Monitoring and Logging. AWS deployments follow a parallel structure. Preview environments materialize on every pull request without additional configuration.
Local development mirrors production closely enough that teams can spin up multi-service environments—complete with real databases and pub/sub—on their laptops with a single command. It's accessible through a dashboard at localhost, which is a small detail that matters when you're trying to close the gap between what developers test locally and what actually runs in production.
Enter the AI Agent
Leap launched on May 31, 2025, and it represents something more ambitious than another code completion assistant. The agent generates entire backend systems: APIs, microservices, database schemas, and the infrastructure declarations required to actually deploy them. It integrates with GitHub and deploys through Encore Cloud to AWS or GCP.
The workflow is opinionated, deliberately so. A developer describes what they want to build. Leap generates the Encore.ts application, including both backend logic and infrastructure declarations. That code goes into version control. From there, Encore Cloud handles provisioning, deployment, and operational visibility.
Unlike the proliferation of "toy" app builders that spit out boilerplate unsuitable for production use, Leap aims at generating apps that slot directly into Encore's monitoring and deployment pipeline. Teams inherit the observability layer without manually instrumenting code or integrating external APM tools—a time sink that anyone who has tried to retrofit tracing into an existing service knows well.
The Pragmatic Choices
Authentication is handled through documented integrations with providers like Clerk and Stytch, rather than building yet another auth system. Encore manages secrets and provides type-safe endpoint protection. The platform's documentation includes tutorials and starter templates, which suggests the founders understand that even consolidated platforms require onboarding friction to be minimal.
Deployment requires a single command. Development environments can run on Encore's infrastructure under fair-use limits, or teams can deploy to their own AWS or GCP accounts through Encore Cloud Pro. The platform handles certificate rotation, IAM role provisioning using AWS IRSA patterns, log aggregation, and metrics collection. Cost monitoring exists for GCP deployments now; AWS support is listed as coming, a detail that may matter to teams already deep in the Amazon ecosystem.
Consolidation Isn't New. This Approach Might Be.
The thesis that backend tooling should feel unified rather than cobbled together isn't novel. Supabase bundles Postgres, auth, storage, and edge functions into an open-source backend-as-a-service. Northflank integrates CI/CD, managed databases, and observability. Railway added an observability dashboard for project-level metrics and logs. AWS Amplify has tightened integration between Cognito, AppSync, and hosting.
What differentiates Encore—at least in theory—is that infrastructure lives in application code while still deploying to your own cloud account. There's no vendor lock-in on compute for production workloads, which addresses one of the persistent concerns with platform-as-a-service offerings. The company advertises incremental adoption: teams can deploy to existing Kubernetes clusters or provision new infrastructure. Migration documentation addresses bringing existing services into Encore's model, though how smooth that process is in practice remains to be seen.
The platform also exposes an MCP server to make services, APIs, and schemas available to AI assistants. It's a bet that agentic workflows will become standard in development teams, rather than a passing novelty.
Marketing materials cite testimonials from Groupon, Bookshop.org, and Pave Bank about reduced DevOps overhead and faster shipping. Specific metrics from those customers aren't detailed publicly, which is typical for early-stage companies still building case study depth.
The Broader Shift
The industry has been moving toward consolidation for a while now, perhaps out of necessity. Datadog launched App Builder in June 2024 to embed actions directly into its monitoring stack—an acknowledgment that context-switching across 550-plus integrated tools creates friction. Dynatrace announced enhancements in January 2026 to unify frontend, backend, AI telemetry, databases, cloud, and mobile into a single developer experience.
The pain point Encore addresses is undeniably real. Qovery's widely-circulated blog post on tool sprawl articulated it plainly: Terraform glue code, Helm configurations, Jenkins pipelines, config drift between environments, and security gaps that emerge when teams stitch together a dozen specialized tools. The argument for unified platforms is straightforward—they abstract Kubernetes complexity while preserving enough control for production use.
But perhaps the more interesting question is whether Encore's specific implementation—infrastructure-from-code combined with AI generation—represents the winning pattern, or just one viable approach among several.
What's Missing From the Picture
Pricing details for Encore Cloud Pro and scaled Leap usage aren't prominently displayed on public pages, though the company offers a free tier for getting started. Enterprise features around SSO, compliance certifications, and multi-region failover exist but lack detailed public documentation—a gap that might matter to larger organizations evaluating whether to adopt the platform.
For teams currently drowning in tooling sprawl, Encore represents a clear hypothesis: the backend development stack should feel more like a single, coherent product than a collection of loosely connected services held together by glue code and hope. Whether that resonates widely enough to displace established patterns of Terraform configs and multi-tool pipelines will depend on a calculation teams make constantly—do they value consolidation more than the flexibility of assembling best-of-breed point solutions?
The answer, as with most platform bets, likely depends on team size, organizational complexity, and how much tolerance remains for managing infrastructure as a separate concern from writing code.
What Eriksson and Kohlberg are banking on is that the tolerance for that complexity has finally run out.
