Heroku review

Heroku is a hosted cloud application platform, not a library or a server you install in your own environment. It runs application processes in dynos on a managed runtime, with Heroku Postgres and a Redis-based key-value store available as managed data services. Deployment and operations sit between your application code and the underlying infrastructure: Heroku handles the runtime environment while your team manages the app and its configuration.

That makes Heroku a platform layer in an application stack. It aims to reduce the infrastructure work involved in building, deploying and scaling web apps, and offers add-ons and buildpacks to extend the runtime. Teams can deploy through GitHub integration or the Heroku CLI, then scale apps up or out across the listed dyno plans. The platform also includes managed inference and agents, MCP hosting, and pgvector for Heroku Postgres. Heroku describes itself as fully managed; the published materials do not establish a self-hosted deployment option.

Heroku’s homepage describes a deployment-first integration surface: the CLI workflow includes git push heroku main, and the AI workflow includes heroku ai:models:create. GitHub integration can deploy a branch on each push, manually or automatically, and create a disposable Review App for a pull request. Heroku Flow supports continuous delivery, while buildpacks customize the language stack and add-ons connect third-party services.

The officially supported languages listed are Node.js, Ruby, Java, PHP, Python, Go, Scala, Clojure and .NET. Heroku also says a third-party buildpack can support any language that runs on Linux. The homepage does not name an API protocol or SDKs, so confirm the required SDK and API support in Heroku’s docs before choosing a language-specific integration. It names Heroku Postgres, the key-value store, managed inference and agents, MCP hosting, and pgvector as platform services. Review the docs for the precise setup and operational requirements of the services your app will use.

Heroku has no free plan in the listed pricing. Eco costs $5 per month and includes 0.5 GB RAM, 1x–4x compute and two process types; it sleeps after 30 minutes of inactivity. Basic is $7 per month, with 0.5 GB RAM and always-on service. For apps that need scaling and monitoring, Standard-1X is $25 per month; Standard-2X is $50 per month and lists 1 GB RAM.

The higher tiers change the bill sharply: Performance-M is $250 per month, while Performance-XL is $750 and Performance-2XL is $1,500. Private and Shield plans start at $125 and $150 per month respectively, and require private networking or provide PCI compliance as listed. Fir compute plans reach $2,400 per month for Compute 32CPU-64GB. These are monthly plan prices, not a complete project estimate: calculate the number and size of processes your app needs, and check current pricing for data services and add-ons. The pricing page lists no annual discount. Confirm billing behavior for scaling, data services and any other usage before production.

Heroku is presented as a fully managed hosted platform. The homepage does not provide a platform license or describe a self-hosting option, so teams that need source access or control over the hosting environment should confirm those points with Heroku before committing. It does describe buildpacks as extensible and says hundreds are available from the community; that does not establish the license of Heroku itself or of each buildpack.

The published materials also do not describe a data-export path or give a migration procedure. Ask how to export application data, database backups and configuration, and whether your chosen add-ons have independent export tools. Apps that depend on dynos, Heroku-specific deployment workflows, add-ons or managed data services will need replacements for those platform components when moving. Keeping application code and data services portable can reduce that work, but the exact migration effort depends on the app and its dependencies.

Heroku lists scaling up or out, autoscaling on several Performance plans, App Metrics, fast rollback and managed data services. Its homepage also says it maintains PCI, HIPAA, ISO and SOC compliance; teams should verify which services, plans and configurations apply to their requirements rather than treating that statement as a blanket guarantee. Private Spaces provide network-isolated runtime environments, and Shield plans list PCI compliance.

Check plan capacity, sleep behavior and the cost of the process count your app needs before launch: Eco sleeps after 30 minutes of inactivity, while Basic is listed as always on. The published pricing lists monthly plans but does not state request limits, rate limits or scaling overage rules; ask Heroku how those affect your workload. Reddit discussions include a Rails developer concerned about dyno prices, a user describing a serious outage, and a thread questioning the platform’s future. These are individual reports, not reliability measurements, but they make continuity, recovery and cost worth validating with your own requirements.