Encore

Encore is a development platform with an unusual pitch: you write your backend in TypeScript or Go using Encore's open-source SDK, declare the infrastructure you need — services, databases, caches, Pub/Sub topics, cron jobs, storage buckets — directly in application code, and Encore's platform provisions matching resources inside your own AWS or GCP account. No Terraform, no YAML glue, no separate infrastructure repository to keep in sync. We sit it in our managed cloud category with the caveat that Encore does not sell you compute: you pay AWS or GCP for the machines, and you pay Encore a platform fee for the automation, environments and tooling wrapped around them.

What Encore actually is

Three layers matter. First, the SDK: an open-source framework (12K+ GitHub stars at the time of writing) where a backend primitive is a typed declaration in code — a service is a module, a database is a declaration, a Pub/Sub subscription is a declaration. Encore parses the application, generates the API documentation and architecture diagrams from it, and runs the whole stack locally with one command, spinning up local implementations of the infrastructure you declared. Second, the automation: on the paid plan Encore analyzes the same code and provisions real resources in your AWS or GCP account — including IAM and networking — then runs CI/CD with GitHub integration, preview environments per pull request, and change-management approvals. Third, the observability: distributed tracing, metrics and a service catalog are generated from the code, so you are not stitching Datadog to a hand-drawn architecture diagram.

The company history is visible in the product. Encore started as founder André Eriksson's side project in 2017 while he was a staff engineer at Spotify, was open-sourced and incorporated in 2021, and is run by a distributed European team with Spotify and Google backgrounds. The roadmap now leans hard toward AI coding agents: an MCP server and guardrails so agents can build, test and deploy against real preview environments. Customer names they publish include Groupon, Playwire, Bookshop.org, Pave Bank and Echo (acquired by Coinbase) — mostly scale-ups, fintechs and teams modernising legacy stacks rather than enterprises with large platform departments.

The architecture claim we care most about: everything runs in your own cloud account. Encore provisions standard AWS and GCP resources, and their documentation describes how to keep running if you stop paying Encore. That is a real escape hatch, and it is the main thing separating this from a PaaS that hosts your workloads on its own metal — see our serverless platforms coverage for that model. When you cancel, your infrastructure stays up; you lose the orchestration, preview environments and tracing, not the production stack.

Pricing: what you really pay

Encore's published plans are unusually itemised. Starter is free forever: local development with automated infrastructure, two free development environments on Encore Cloud, 1M tracing events per month with 7-day retention, CI/CD and automated testing, community support on Discord. Fair-use limits apply to Encore Cloud environments — 100,000 requests per day, 100,000 Pub/Sub messages per day, 1 GB database storage and 1 GB object storage per application.

Pro is $49 per member per month, plus $99 per AWS/GCP environment per month (metered hourly at $0.136/hour), plus $2.50 per resource per environment per month (metered at $0.0034/hour), plus $1.20 per 1M tracing events beyond the 20M included. A "resource" is anything you declare in code — services, databases, cron jobs, caches, storage buckets, Pub/Sub topics and subscriptions, secrets. Crucially, cloud usage is billed directly by AWS or GCP in your own account; Encore's fee is the platform only. There is a 14-day free trial of Pro with no credit card, and metering means a throwaway environment spun up for a week costs a fraction of the monthly rate. Enterprise is custom pricing: SSO/SAML, custom RBAC, SOC 2 compliance assistance, self-hosted preview environments, a technical account manager, uptime guarantees and custom support SLAs, invoicing by bank transfer. Startups and open source projects can apply for free credits toward Pro.

Performance & infrastructure

We have not benchmarked Encore — our benchmark programme measures VPS and bare metal, not application frameworks — so this is analysis, not test data. The performance story is really a cost and utilisation story: your workloads run on ordinary AWS and GCP services (their FAQ is explicit that you pay the cloud provider for compute, storage and data), so raw performance is whatever instance type you provision. What Encore changes is the shape of the bill and the speed of iteration: environment and resource metering is hourly, preview environments exist only while the pull request does, and there is no platform-side compute markup to pay. On the local side, the one-command local stack with real service primitives is where we would expect the biggest time saving — no hand-built docker-compose estate drifting out of sync with production.

Reliability of the control plane is the open question. Encore Cloud hosts the orchestration, tracing and preview environments, and uptime guarantees are listed only as an Enterprise feature. Since your production infrastructure runs in your own account, a platform outage degrades tooling and deployments rather than taking down your service — a better failure mode than a hosted PaaS, though a failed deploy window is still a failed deploy window. Their own comparison pages are careful about this distinction, and we credit them for it.

Where it falls short

The obvious cost: Pro pricing stacks. A five-person team with two cloud environments and twenty declared resources is paying several hundred dollars a month on top of the AWS bill, before a single tracing event overage. That is defensible if Encore replaces a DevOps hire; it is hard to justify for a two-person side project, and the free Starter plan cannot deploy to your own cloud at all — it is local development plus Encore Cloud only.

Second, the framework tax. Encore's patterns — how services declare each other, how APIs are typed, how infrastructure is expressed — are Encore's own. The SDK is open source and the generated infrastructure is standard, which is more of an exit than most platforms offer, but your application code is written against Encore's abstractions. Teams that need deep control of IAM boundaries, exotic AWS services not yet modelled in the SDK, or a bespoke networking topology will hit the edges of what "infrastructure from code" can express; there is a reason the Enterprise tier sells custom network security and RBAC rather than pretending the defaults fit everyone. Terraform is verbose, but nothing is hidden from you.

Third, platform maturity. Encore is a venture-backed company of roughly startup size, and it is currently repositioning around AI coding agents. That is a sensible bet, but it means roadmap attention is a moving target. Languages beyond TypeScript and Go are not at feature parity, and the tracing and observability tooling, while convenient, will not replace a mature APM stack for teams with serious compliance or retention requirements — retention on Pro is 30 days.

Who should pick this

Our pick for seed-to-Series-B teams building on AWS or GCP who do not want to hire a platform engineer just yet: the automatic provisioning, per-pull-request environments and generated docs cover most of what a small team would otherwise glue together with Terraform, GitHub Actions and an APM trial. It also fits regulated and fintech workloads specifically because the infrastructure lands in your own account — your compliance story is your cloud account's compliance story. Platform teams modernising a legacy monolith into services are the other sweet spot; the service catalog and change-management tooling earn their keep there.

We would not pick Encore for a static site, a single CRUD app on one small VM, or a team that already runs mature Terraform and CI/CD — the marginal gain over a working setup is small, and the per-member and per-resource fees are real. If you want a platform to host the workloads for you and never think about AWS, a hosted PaaS is simpler and often cheaper at small scale. If you want maximum infrastructure control with code, plain Terraform or Pulumi keeps every decision in your hands.

Key takeaways

• Infrastructure from code: declare services, databases, caches, Pub/Sub and cron in TypeScript or Go; Encore provisions them in your own AWS or GCP account.
• Transparent, itemised pricing: Starter free (local dev plus two Encore Cloud environments), Pro at $49 per member per month plus $99 per cloud environment and $2.50 per resource per environment, metered hourly.
• Your cloud bill stays with AWS or GCP — Encore charges platform fees only, and the SDK is open source (12K+ GitHub stars) with a documented migration path.
• No hands-on performance claims from us — our benchmarks cover servers, not frameworks; production speed is whatever you provision.
• Caveats: per-member and per-resource fees add up, application code is written against Encore's abstractions, and uptime SLAs are Enterprise-only.
• Best fit: small product teams on AWS/GCP who want platform-engineering outcomes without a platform team; overkill for one-VM projects.

Sources

• https://encore.dev/pricing
• https://encore.dev/about