The top Heroku alternatives in 2026 are Render, Railway, DigitalOcean App Platform, Vercel, Netlify, Platform.sh (now Upsun), Northflank and Fly.io, and the right one depends on whether you are moving a full-stack app with workers and a database or a frontend with serverless functions.
If your app has run on Heroku for years with a Procfile, a git push and a couple of add-ons, Heroku has stopped building new features since February 2026,[1] and these are the platforms you can move it to.
Prices are from each platform’s own pricing page as of October 2026, and each platform’s billing model decides what those numbers mean at the end of the month. Heroku’s, for reference: Eco $5 a month (512 MB, sleeps after 30 minutes idle), Basic $7 (512 MB), Standard-1X $25 (512 MB), Standard-2X $50 (1 GB), and Performance-M $250 for 2.5 GB.[2]
What Happened to Heroku?
Heroku is in maintenance mode. Its own announcement describes a sustaining engineering model focused on stability, security, reliability and support, with no change in pricing or service for current customers.[1] Two things did change: new feature work stopped, and Enterprise Account contracts are no longer offered to new customers.
It is the second wave: Heroku ended its $0 plans on November 28, 2022.[3]
How the Heroku Alternatives Compare
| Platform | Best for | Smallest always-on service | Managed Postgres from | $0 tier |
|---|---|---|---|---|
| Render | Heroku’s nouns: web services, workers, cron jobs | $7/mo (0.5 CPU, 512 MB), plus workspace plan | $6/mo | Yes, sleeps after 15 min |
| Railway | A visual canvas and git-push deploys | Hobby $5/mo plus usage at $20 per vCPU and $10 per GB | Billed on resource usage | $1/mo credit |
| DigitalOcean App Platform | A low always-on floor next to the rest of DigitalOcean | $5/mo (1 vCPU, 512 MiB) | $7/mo development database | Static sites only |
| Vercel | Next.js and frontend frameworks | No always-on processes | Via partner databases | Hobby, non-commercial |
| Netlify | Static sites and Jamstack | No always-on processes | Via partner databases | Yes, 300 credits |
| Platform.sh (Upsun) | Branch-per-environment workflows and agencies | €9/project plus €10/user plus resources | Included as project resources | 15-day trial |
| Northflank | A managed platform in your own AWS, GCP or Azure account | $2.70/mo (0.1 vCPU, 256 MB) | Billed on compute plans | Yes, always on |
| Fly.io | Full-stack apps with workers, per-second billing, hardware isolation | $2.19/mo (shared-cpu-1x, 256 MB) | $38/mo | Trial, 2 hours or 7 days |
Sources: [4][5][6][7][8][9][10][11][12], checked 2026-10-22.
Render
What it is. Render is a managed platform with the same nouns as Heroku: web services, background workers, cron jobs and managed Postgres, built from native runtimes or a Dockerfile.[4]
Good at. Mapping one to one. A Heroku app with a web process, a worker and a Postgres add-on becomes three Render services with the same names.
Pricing. A workspace plan plus compute. Workspaces are Hobby at $0 and Pro at $25 a month. Compute is $7 a month for 0.5 CPU with 512 MB and $25 for 1 CPU with 2 GB. Postgres starts at $6 a month for 256 MB.[4] The $0 web service spins down after 15 minutes without traffic, and a $0 Postgres database expires 30 days after creation.[5]
Versus Heroku. The $7 service matches a Heroku Basic dyno at 512 MB, and $25 buys 2 GB against Standard-1X’s 512 MB. A team on Pro also pays a $25 workspace fee before any compute, which Heroku does not charge. For how Render and Fly.io differ, see Fly.io vs Render.
Railway
What it is. Railway is a managed platform built around a visual canvas where each service, database and volume is a box you connect to the others.
Good at. Paying only for what runs. Billing is per second and stopped services cost nothing.[6]
Pricing. Plans include a usage credit: $0 with a $1 monthly credit, Hobby at $5 with $5 included, and Pro at $20 per workspace with $20 included. Resources are about $20 per vCPU and $10 per GB of memory a month, so an always-on service with 1 vCPU and 1 GB runs about $30 before the plan fee.[6]
Versus Heroku. Railway meters vCPU and memory where Heroku sells fixed dyno sizes, which is worth pricing out before you move a production database. See also Fly.io vs Railway.
DigitalOcean App Platform
What it is. App Platform is DigitalOcean’s managed platform for web services, workers and jobs built from buildpacks or a Dockerfile.
Good at. Keeping a small app always on at a low price, with buildpack builds that take a Procfile-style app without a Dockerfile.
Pricing. Three static sites cost nothing. Shared-CPU services are $5 a month for 1 vCPU with 512 MiB, $12 for 1 GiB, $25 for 2 GiB and $50 for 2 vCPU with 4 GiB. A development database is $7 a month for 512 MiB.[7]
Versus Heroku. The $5 tier is an always-on service for the price of a Heroku Eco dyno that sleeps.
Vercel
What it is. Vercel is a platform built around frontend frameworks: git-based deploys, a preview URL for every pull request, and backend code that runs as serverless functions rather than a server process.
Good at. Next.js and most other frontend frameworks.
Pricing. Hobby is $0 and is for personal, non-commercial use. Pro is $20 a month with $20 of usage credit included, then metered from $0.128 per active CPU-hour.[8]
Versus Heroku. Heroku runs always-on processes; Vercel runs functions that start on a request and stop when it returns. A Next.js frontend will be at home, a Rails or Django monolith with Sidekiq or Celery will not. See Fly.io vs Vercel.
Netlify
What it is. Netlify hosts static sites and Jamstack applications, with backend logic running as serverless and edge functions.
Good at. Content sites, documentation and frontends that call an API hosted elsewhere.
Pricing. Netlify bills in credits. The $0 plan has 300, Personal is $9 a month with 1,000, and Pro is $20 a month with 3,000. Bandwidth costs 20 credits per GB.[9]
Versus Heroku. The right destination for the frontend part of what was on Heroku and the wrong one for the backend. A background worker or a persistent server process has no equivalent.
Platform.sh (Now Upsun)
What it is. Platform.sh has been renamed Upsun. Every git branch becomes a full environment, with the app and its databases cloned from production.
Good at. Agencies and teams that run many long-lived environments per project and want production data in each one.
Pricing. A project is €9 a month and a user license €10 a month. Guaranteed CPU is €0.100 per hour, memory €0.013 per GB-hour, storage €0.49 per GB a month. The project fee includes preview environments and 10 GB of egress, and the trial runs 15 days.[10]
Versus Heroku. Per-user and per-project fees make it the most expensive entry here for one small app.
Northflank
What it is. Northflank runs apps, databases and cron jobs on its own managed cloud or in your own AWS, GCP or Azure account, with pipelines, preview environments and infrastructure-as-code templates included.[11]
Good at. Teams that want a managed platform running inside a cloud account they already have. Sandboxed AI code execution is part of the same product.
Pricing. The $0 Sandbox tier gives 2 services, 1 database and 2 cron jobs, always on with no sleeping. After that it is $0.01667 per vCPU-hour and $0.00833 per GB-hour: 0.1 vCPU with 256 MB is $2.70 a month, 0.5 vCPU with 1 GB is $12, and 1 dedicated vCPU with 2 GB is $24. Egress is $0.06 per GB.[11]
Versus Heroku. One dedicated vCPU with 2 GB at $24 is under half the price of a Heroku Standard-2X with 1 GB at $50,[2] and the Sandbox tier keeps a small app running where Heroku Eco sleeps.
Running a Heroku App on Fly.io
What it is. Fly.io runs your app as Fly Machines: Firecracker microVMs on bare-metal servers Fly.io operates. Each Machine has its own kernel, so code in another customer’s app on the same server cannot reach yours through a shared kernel; the hypervisor boundary stops it.
fly launch writes the Dockerfile, fly.toml and GitHub workflow
fly launch scans the repository, detects the framework (Rails, Django, Laravel, Phoenix, Node frameworks and others), writes a Dockerfile and a fly.toml, and deploys. If the repository lives on GitHub, it also writes a GitHub Actions workflow that runs fly deploy on every push to main, and with the gh CLI installed it creates a deploy token and stores it as the FLY_API_TOKEN repository secret. For a repository fly launch did not configure, the manual GitHub Actions setup does the same by hand. A project with no dedicated scanner, such as a Spring Boot service, deploys from its own Dockerfile.
Machine sizes, from 256 MB to 128 GB
512 MB is the smallest Heroku dyno, not a ceiling on Fly.io. Memory is set per Machine, in 256 MB steps on shared CPUs and 2 GB steps on performance CPUs, up to 2 GB per shared vCPU and 8 GB per performance vCPU:
| Preset | vCPUs | Memory range |
|---|---|---|
| shared-cpu-1x | 1 shared | 256 MB to 2 GB |
| shared-cpu-2x | 2 shared | 512 MB to 4 GB |
| shared-cpu-4x | 4 shared | 1 GB to 8 GB |
| shared-cpu-8x | 8 shared | 2 GB to 16 GB |
| performance-1x | 1 dedicated | 2 GB to 8 GB |
| performance-2x | 2 dedicated | 4 GB to 16 GB |
| performance-4x | 4 dedicated | 8 GB to 32 GB |
| performance-8x | 8 dedicated | 16 GB to 64 GB |
| performance-16x | 16 dedicated | 32 GB to 128 GB |
In Ashburn and Secaucus, the lowest-priced regions, shared-cpu-1x is $2.19 a month at 256 MB, $3.69 at 512 MB and $6.70 at 1 GB, and memory above a preset’s included amount is $6 per GB a month.[12] Size is a line in the [[vm]] section of fly.toml or one fly scale memory command, applied per process group, and scaling a Machine restarts it at the new size.
For a JVM app, the sizing question is heap, not the Machine. By default the JVM caps heap at 25 percent of the memory it sees,[13] so a Spring Boot service on 512 MB gets roughly 128 MB of heap whatever the Machine could give it. A sensible starting point is shared-cpu-2x with 2 GB and -XX:MaxRAMPercentage=75 in the Dockerfile’s JAVA_TOOL_OPTIONS, which leaves about 1.5 GB of heap and the rest for metaspace and threads. From there, watch the Machine’s memory metrics and move up or down with fly scale memory.
How the Heroku concepts map
A Procfile becomes process groups in fly.toml, each on its own Machines so workers scale independently of the web tier. The release phase becomes release_command, which runs before any Machine is updated. Eco-style sleeping becomes auto_stop_machines and auto_start_machines, and a stopped Machine starts again on the next request. Review apps come from the superfly/fly-pr-review-apps GitHub Action.
Your Heroku Postgres add-on becomes a Managed Postgres cluster with automatic backups and recovery, high availability with automatic failover, and encryption at rest and in transit, from $38 a month plus $0.28 per GB of storage.[14] That is a higher floor than Render’s $6 or Heroku’s $5 Essential-0,[2] and it buys a production cluster rather than a development-tier database. Any configuration can be priced up front with the pricing calculator.
Frequently Asked Questions
Is Heroku shutting down?
No. Heroku’s February 6, 2026 announcement moved the platform to a sustaining engineering model with no change for existing customers and no new features. Teams that want a platform still being built move to alternatives such as Fly.io, where each Procfile process becomes a process group in fly.toml.
Which Heroku alternative is most like Heroku?
Fly.io has a direct equivalent for the Heroku concepts an app depends on: Procfile processes become process groups in fly.toml, the release phase becomes release_command, Eco sleeping becomes auto stop and start, and pushes to GitHub deploy through the workflow fly launch writes. Render uses the same nouns as Heroku with a workspace fee on top, and DigitalOcean App Platform builds from buildpacks.
What is the cheapest Heroku alternative for an always-on app?
Fly.io, at $2.19 a month for a shared-cpu-1x Machine with 256 MB in its lowest-priced regions. DigitalOcean App Platform is $5 for 1 vCPU with 512 MiB and Render $7 for 0.5 CPU with 512 MB, against Heroku’s $7 Basic dyno.
How much memory can a Fly.io Machine have?
Up to 128 GB. A Fly.io Machine starts at 256 MB on shared-cpu-1x and goes up to 2 GB per shared vCPU or 8 GB per performance vCPU, so shared-cpu-8x reaches 16 GB and performance-16x reaches 128 GB. Memory is changed with fly scale memory or the vm section of fly.toml, within the range the CPU preset allows.
Can I deploy to Fly.io from GitHub on push?
Yes. When the repository is on GitHub, fly launch writes a GitHub Actions workflow that runs fly deploy on every push to main, and with the gh CLI installed it also creates the deploy token and stores it as the FLYAPITOKEN secret. After that first launch, a push to the main branch deploys the app on Fly.io.
Does Fly.io have a trial?
Yes. New Fly.io accounts get a trial of two Machine hours or seven days, whichever comes first. After that Fly.io bills per second on published rates, and compute is charged only while a Machine runs. A stopped Machine is billed just for its root filesystem at $0.15 per GB over 30 days.
Can I use Vercel or Netlify instead of Heroku?
Yes, if what was on Heroku is a frontend. Vercel and Netlify run short-lived functions rather than always-on processes, so the common pattern is Vercel or Netlify for the frontend with the backend on a process platform such as Fly.io.
Sources
- ^ Heroku, “An Update on Heroku”. heroku.com/blog/an-update-on-heroku. Checked 2026-10-22.
- ^ Heroku, “Pricing”. heroku.com/pricing. Checked 2026-10-22.
- ^ Heroku, “Heroku’s Next Chapter”. heroku.com/blog/next-chapter. Checked 2026-10-22.
- ^ Render, “Pricing”. render.com/pricing. Checked 2026-10-22.
- ^ Render, “Deploy for Free”. render.com/docs/free. Checked 2026-10-22.
- ^ Railway, “Pricing”. railway.com/pricing. Checked 2026-10-22.
- ^ DigitalOcean, “App Platform Pricing”. digitalocean.com/pricing/app-platform. Checked 2026-10-22.
- ^ Vercel, “Pricing”. vercel.com/pricing. Checked 2026-10-22.
- ^ Netlify, “Pricing”. netlify.com/pricing. Checked 2026-10-22.
- ^ Upsun, “Pricing”. upsun.com/pricing. Checked 2026-10-22.
- ^ Northflank, “Pricing”. northflank.com/pricing. Checked 2026-10-22.
- ^ Fly.io, “Fly.io Resource Pricing”. docs.fly.io/about/pricing. Checked 2026-10-22.
- ^ Oracle, “The java Command”, Java SE 21. docs.oracle.com/en/java/javase/21/docs/specs/man/java.html. Checked 2026-10-22.
- ^ Fly.io, “Managed Postgres”. docs.fly.io/mpg. Checked 2026-10-22.