Best Heroku Alternatives in 2026
An honest comparison of the best Heroku alternatives in 2026 (Railway, Render, Fly.io, DigitalOcean App Platform, Vercel, and Northflank) on pricing, Heroku migration effort, PostgreSQL support, and deployment simplicity.
Last updated: November 2026
Your app runs on Heroku, and in February 2026 Salesforce announced that Heroku is "transitioning to a sustaining engineering model focused on stability, security, reliability, and support." Enterprise Account contracts are no longer offered to new customers. Existing apps keep running and existing contracts are honored, but if you're planning the next few years of your app, you're probably weighing where it should live.
On Railway, you can keep the git-push workflow you're used to. Connect your GitHub repository, paste in your Heroku config vars, and restore your Postgres data. Your web process, workers, and databases then run side by side in one project.
This guide compares Railway with Render, Fly.io, DigitalOcean App Platform, Vercel, and Northflank. For each one, it covers how pricing works, how much work the move from Heroku takes, how PostgreSQL is handled, and how you deploy. It ends with step-by-step instructions for moving an app to Railway.
How do Heroku alternatives compare?
A typical Heroku app is a web process, maybe a worker, and a Postgres database, so start with how each platform runs those and what it costs. The table shows where each platform fits and where its paid plans start. Your total also depends on the resources and services you use.
| Platform | Good fit | Moving from Heroku | Paid starting point |
|---|---|---|---|
| Railway | Heroku-style apps with several services and databases in one project | Procfile support, paste config vars in .env format, dump and restore Postgres | Hobby $5/month, including $5 of usage |
| Render | Teams that want fixed instance prices and managed Postgres | Manual, with a documented guide | Web services from $7/month; Pro workspace $25/month |
| Fly.io | Apps that need to run close to users in many regions | CLI-driven, with a documented guide | Usage-based machines from about $2.19/month; Managed Postgres from $38/month |
| DigitalOcean App Platform | Teams already on DigitalOcean | Manual, with a documented guide | Containers from $5/month; Managed PostgreSQL from $15.15/month |
| Vercel | Frontends and Next.js apps | Backends need refactoring into serverless functions | Pro $20/month per developer seat, including $20 of usage |
| Northflank | Teams that must deploy into their own cloud account | Manual, with a documented guide | Free Sandbox; pay as you go from $0.01667 per vCPU-hour |
Sources, prices as of November 2026: Railway pricing, Render pricing, Fly.io pricing, DigitalOcean App Platform pricing and Managed Databases pricing, Vercel pricing, and Northflank pricing.
Why are teams moving off Heroku?
Heroku's move to a sustaining engineering model means it's focusing on stability rather than new features. For many teams, that turns some long-standing limits into permanent ones.
Heroku has no persistent volumes. Anything your app writes to the local filesystem is lost on the next deploy. There's also no native multi-region support, so serving users globally means separate apps in each region behind an external load balancer.
Each Heroku app also stands alone. There's no project that groups an API, a worker, and a frontend, and each app has its own isolated config vars, so a secret shared by three services is set three times. Private networking is a paid add-on, available only on enterprise plans.
Zero-downtime deploys take some assembly. You enable Preboot so new dynos start serving traffic before old ones stop, use a release phase for backward-compatible migrations, and handle SIGTERM for graceful shutdown. Heroku has metrics and logs, but no built-in HTTP health checks, so you add your own health-check endpoint and external monitoring to catch a bad deploy.
If your app runs well on Heroku and none of these limits affects you, you don't have to move now. Existing apps keep running, and existing contracts are honored.
How do the platforms compare feature by feature?
Railway, Render, and Fly.io support persistent volumes, while Heroku, DigitalOcean App Platform, and Vercel don't. Vercel is the only platform here that runs your code as serverless functions rather than long-running servers. In the table, ✅ means full support, ⚠️ means partial support or a workaround, and ❌ means not supported.
| Feature | Railway | Render | Fly | DigitalOcean | Vercel | Northflank | Heroku |
|---|---|---|---|---|---|---|---|
| DEPLOYMENT | |||||||
| Deployment Model | Long-running servers | Long-running servers | Lightweight VMs | Long-running servers | Serverless functions | Long-running containers on Kubernetes | Long-running servers |
| Docker Support | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ❌ No | ✅ Yes | ✅ Yes |
| Source Code Deploy | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes (Dockerfile or Buildpacks) | ✅ Yes |
| Multi-Service Projects | ✅ Yes | ✅ Yes | ❌ No | ✅ Yes | ❌ No (one-to-one) | ✅ Yes | ❌ No |
| INFRASTRUCTURE | |||||||
| Runs On | Own hardware | AWS/GCP | Own hardware | Own hardware | AWS (serverless) | Northflank's cloud, or your own cloud account (BYOC) | AWS |
| Max Memory | Plan-based | Instance-based | Configurable | Instance-based | 4GB | Plan-based | Instance-based |
| Execution Limits | None | None | None | None | 13.3 min max | None | None |
| Cold Starts | No | No | No | No | Yes (however several optimizations exist to reduce them) | No | No |
| Persistent Storage via volumes | ✅ Yes | ✅ Yes | ✅ Yes | ❌ No | ❌ No | ✅ Yes | ❌ No |
| DATABASES & STORAGE | |||||||
| Database Support | ✅ One-click deploy any open-source database | ✅ Native | ✅ Native | ✅ Native | Via marketplace | ✅ Managed addons (Postgres, MySQL, MongoDB, Redis, and more) | ✅ Native (via add-ons) |
| SCALING | |||||||
| Vertical AutoScaling | ✅ Automatic | ⚠️ Manual/threshold | ⚠️ Manual/threshold | ⚠️ Manual/threshold | ✅ Automatic | ⚠️ Manual (resize the plan) | ⚠️ Manual/threshold |
| Horizontal Scaling | ✅ Yes (By deploying replicas) | ✅ Yes (Configure min and max number of concurrent instances) | ✅ Yes (By deploying fly-autoscaler) | ✅ Yes (Configure min and max number of concurrent instances) | ✅ Yes | ✅ Yes (autoscaling on paid plans) | ✅ Yes (Configure min and max number of concurrent instances) |
| Multi-Region Support | ✅ Native | ❌ No (requires manual setup) | ✅ Native | ❌ No (requires manual setup) | ✅ Yes | ⚠️ One region per project | ❌ No (requires manual setup) |
| PRICING | |||||||
| Pricing Model | Usage-based (Active compute time + resources used) | Instance-based | Machine state-based | Instance-based | Usage-based (Active compute time + resources used) | Usage-based (provisioned resources, billed per second) | Instance-based |
| Billing Factors | Active compute time × size | Fixed monthly per instance. When scaling horizontally it's instance size x total running time | Running time + CPU type | Fixed monthly per instance | CPU time + memory + invocations | Configured vCPU + memory × running time | Fixed monthly per instance |
| Scales to Zero | ✅ Supported via app sleeping | ❌ No | ✅ Supported via autostop | ❌ No | ✅ Yes | ⚠️ Manual (scale to 0 or pause) | ❌ No |
| CI/CD & ENVIRONMENTS | |||||||
| GitHub Integration | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes |
| PR Preview Environments | ✅ Yes | ✅ Yes | ⚠️ Not supported out of the box. Requires setting up a CI/CD pipeline | ❌ No | ✅ Yes | ✅ Yes | ✅ Yes |
| Environment Support | ✅ Built-in | ✅ Built-in | ⚠️ Separate orgs | ⚠️ Separate projects | ✅ Built-in | ✅ Built-in | ✅ Built-in |
| Instant Rollbacks | ✅ Yes | ✅ Yes | ❌ No | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes |
| Pre-Deploy Commands | ✅ Yes | ✅ Yes | ⚠️ Manual when setting up a deployment pipeline | ✅ Yes | ✅ Yes | ✅ Yes (job step in a workflow) | ✅ Yes |
| OBSERVABILITY | |||||||
| Built-in Monitoring | ✅ Yes | ✅ Yes | ✅ Yes (Prometheus) | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes |
| Integrated Logs | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes |
| DEVELOPER TOOLS | |||||||
| Infrastructure as Code | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes (templates, GitOps) | ✅ Yes |
| CLI Support | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes |
| SSH Access | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ❌ No | ✅ Yes (shell via CLI or UI) | ✅ Yes |
| Webhooks | ✅ Yes | ✅ Yes | ❌ No | ❌ No | ✅ Yes | ✅ Yes | ✅ Yes |
| NETWORKING | |||||||
| Custom Domains | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes |
| Managed TLS | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes |
| Private Networking | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ❌ No | ✅ Yes | ⚠️ Paid add-on |
| Health Checks | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes |
| ADDITIONAL FEATURES | |||||||
| Native Support for Cron Jobs | ✅ Yes | ✅ Yes | ❌ No | ❌ No | ✅ Yes | ✅ Yes | ✅ Yes |
| Shared Variables | ✅ Yes | ✅ Yes | ⚠️ Manual | ✅ Yes | ⚠️ Within project | ✅ Yes (secret groups) | ❌ No |
Where can you move a Heroku app?
Railway: Heroku's workflow, with every service in one project

Railway suits teams that want to keep deploying the way they do on Heroku while running several services and databases together. Most of what you use on Heroku has a direct equivalent: GitHub autodeploys, Docker images, PR environments, instant rollbacks, metrics and logs, a CLI, pre-deploy commands, custom domains with managed TLS, SSH, and webhooks.
You can deploy from a GitHub repository, point at a Docker image, or run railway up from your terminal. You don't need a Dockerfile or a cluster. Railway's Heroku migration guide covers the move, and you can paste the output of heroku config --shell straight into a service's Raw Editor.
Railway reads Procfiles, but only one process type per service. If your Heroku app runs web and worker dynos, create one Railway service for each and set its start command.
Those services live together on the project canvas, along with their databases. You can see everything that's running in one view and how the pieces connect. Services share variables and reference each other's credentials, so you set a secret once instead of once per app. Public and private networking are included at no extra cost.
Deploy everything in one placeYou can create an isolated environment in one click, or turn on automatic PR environments so each pull request gets its own copy of the project.
Create isolated environments in one clickEach service can use CPU and memory up to your plan's limits, so you don't pick a dyno size or tune scaling thresholds. To scale out, add replicas. Railway spreads public traffic randomly across replicas within each region, and each replica runs with your plan's full resource limits. Place replicas in more than one region and Railway routes each request to the nearest one, which on Heroku takes an external load balancer. Railway runs on hardware it owns and operates in its own data centers.
Creating replicas for horizontal scalingPricing follows what each service uses rather than the instance size you reserve:
Active compute time x compute size (memory and CPU)
If you run several replicas, each is billed for its own active compute time. The Hobby plan is $5 a month and Pro is $20 a month per workspace with unlimited seats, and each plan's fee includes the same amount of usage. New accounts get a one-time $5 trial credit for 30 days, with no credit card required. Bills follow your traffic, so they can change from month to month. Set usage alerts so you see spending rise. If you also set a hard usage limit, leave room for normal traffic changes: hitting it takes workloads offline.
App sleeping can reduce idle usage by suspending a service after 10 minutes without outbound traffic. The next request wakes it, so that visitor waits through a cold boot. We'd leave sleeping off for a production web service and use it for staging environments, internal tools, and low-traffic apps.
One-click Postgres runs in your project as a single node by default. You can convert it to high availability with automatic failover, and turn on backups and point-in-time recovery. Before you move production data, turn on backups, and HA if your app needs it. Major version upgrades run in place, with a preflight check and one-click revert. Railway doesn't add extra extensions to its Postgres templates, and if you customize the template's image, start command, or variables, the database becomes yours to manage.
Other databases deploy the same way: MySQL, ClickHouse, Timescale, Redis, Dragonfly, Chroma, Weaviate, and MongoDB, among others. Browse the storage templates for the full list. Because databases live in the same project as your services, you manage them from one dashboard instead of across Heroku add-on providers.
You can attach a volume to any service, and the data on it persists across deploys. Cron jobs run natively on any schedule, where Heroku Scheduler offers three frequencies: every 10 minutes, hourly, or daily. Set a healthcheck path and Railway switches traffic to a new deployment only after it responds. Infrastructure as code gives you programmatic control over your resources.
Integrated metrics and logs show how each deployment is performing without an external tool.
Observability Dashboard
Railway is SOC 2 Type II and SOC 3 certified, with reports available through the Trust Center. For HIPAA workloads, Railway offers a BAA as an add-on above a monthly spend threshold, under a shared responsibility model.
Render: a Heroku-style PaaS with fixed instance prices

Render suits teams that want a Heroku-like PaaS with fixed per-instance pricing and fully managed Postgres. Like Railway, it groups services under one project, such as a frontend, APIs, and databases.
You pick an instance size with a fixed monthly price, on top of a workspace plan: Hobby (free), Pro ($25 a month), or Scale ($499 a month), plus compute. Hobby is limited to one seat, so a team needs a paid plan, and horizontal autoscaling and environments also require one. Paid web services start at $7 a month with 512 MB of memory. Free instances spin down after 15 minutes without traffic and take about a minute to wake.

Render Instances
To scale vertically, you move to a larger instance size. To scale horizontally, you set the instance count yourself or define a minimum and maximum and let Render autoscale on CPU or memory targets. Render has no native multi-region support, so serving users globally means separate instances in each region behind an external load balancer, as on Heroku. Render runs on AWS and GCP.
Render's Heroku migration guide is manual. You create datastores, create services from your Procfile, and copy config vars by hand. Services don't share environment variables by default, so use environment groups or secret files for shared values. Render also offers hands-on help for larger migrations.
Render Postgres is managed, with backups and recovery, high availability, and read replicas. You move data with Heroku's latest.dump and pg_restore. Free Postgres databases expire 30 days after creation.
Deploys are Git-connected from the dashboard, with Blueprints (render.yaml) for infrastructure as code. Render also supports volumes, native cron jobs, healthchecks, and static sites served from a global CDN.
Fly.io: machines you place close to users

Fly.io suits latency-sensitive apps that need to run in many regions, for teams comfortable working from a CLI. Your code runs on Fly Machines, lightweight virtual machines on hardware Fly owns and operates, with native multi-region deployment.
Each machine has a set amount of CPU and memory. You choose a preset size or configure them separately. Shared CPUs guarantee 6% of CPU time with bursting and are throttled under heavy use. Performance CPUs are dedicated and aren't throttled.
To scale, you pick a larger machine through the CLI or API, or run more machines. You can set the machine count yourself, or create a pool of machines in one or more regions and let Fly's Proxy start and suspend them based on incoming requests. Metrics-based autoscaling isn't built in. You can deploy fly-autoscaler, which polls metrics you define and creates, deletes, starts, or stops machines.

Pricing is pay as you go, billed per second for running machines, with no monthly platform fee. Charges depend on machine state and CPU type. A running machine is billed for compute whether or not it's busy, and a stopped machine is billed only for its root file system storage. The smallest always-on machine (shared-cpu-1x, 256 MB) costs about $2.19 a month. New organizations don't get a free tier; new accounts get a $1 trial credit for 7 days. Reserved compute blocks are paid annually up front, in exchange for monthly credits that don't roll over.

Fly's Heroku migration guide works through flyctl. You launch the app, create a Managed Postgres cluster, and pipe a pg_dump of your Heroku database into it through a local proxy. Expect some harmless errors about Heroku-specific extensions. Managed Postgres starts at $38 a month for the Basic plan, plus storage. It runs separately from your apps, so deleting an app doesn't delete its database.
You deploy with fly launch and fly deploy, configured in fly.toml. PR preview environments and instant rollbacks aren't built in, so you'd set them up with a CI tool such as GitHub Actions. For separate development, staging, and production environments, you create an organization for each and link them to a parent organization for billing.
Fly collects metrics from every app through a managed Prometheus service based on VictoriaMetrics, covering HTTP responses, TCP connections, memory, CPU, disk I/O, network traffic, and filesystem use. The dashboard's Metrics tab shows the basics. For detailed dashboards, Fly runs a managed Grafana instance at fly-metrics.net that you query with MetricsQL, and you can connect other tools through the Prometheus API.

Fly supports volumes and healthchecks. It doesn't offer webhooks, which Heroku does.
DigitalOcean App Platform: a classic PaaS for teams already on DigitalOcean

DigitalOcean App Platform suits teams already using DigitalOcean that want a familiar, Heroku-style PaaS with predictable instance pricing. It covers much of what Heroku does: deploys from Docker images or GitHub, long-running servers, public and private networking, instant rollbacks, metrics and logs, a CLI, pre-deploy commands, custom and wildcard domains with managed TLS, and SSH access.
App Platform pricing is instance-based. Up to three static-site apps are free, and paid containers start at $5 a month (shared 1 vCPU, 512 MiB). As with dynos, you pay for the instance size you choose whether or not the app uses all of it. Horizontal autoscaling uses minimum and maximum instance counts and CPU or memory thresholds that you tune.

DigitalOcean's Heroku migration guide maps Heroku processes, the release phase, and Scheduler tasks to App Platform components. You recreate config vars, or use Bindable Variables for values shared between services. DigitalOcean's database migration tool can keep your Postgres data in sync during the move. Pipelines and Review Apps aren't migrated.
For Postgres, you can use a $7-a-month development database, which is PostgreSQL only, isn't backed up by default, and is deleted with its app. For production, Managed PostgreSQL clusters start at $15.15 a month for a single node with 1 vCPU and 1 GiB of RAM.
You deploy from GitHub or a Docker image through the dashboard, CLI, or an app spec. A project can hold several services, such as a frontend, APIs, and databases, and health checks let you deploy without downtime.

App Platform runs in one region per app. A global deployment means separate instances in each region behind an external load balancer, as on Heroku. It has no persistent volumes, so anything written to the local filesystem is lost on redeploy. There are no built-in environments, so development, staging, and production are separate projects. PR preview environments need a CI tool such as GitHub Actions, and App Platform doesn't offer webhooks.
Vercel: serverless hosting for frontends and Next.js

Vercel suits frontend and Next.js apps. It isn't a drop-in replacement for Heroku backends, workers, or long-running jobs, because it doesn't run Docker images, containers, or long-running processes.
With framework-defined infrastructure, Vercel reads your application code at build time and creates the infrastructure it needs. Server-side code is deployed as serverless functions. Each incoming request gets a function instance, and Fluid compute lets one instance handle concurrent requests. Functions scale down to zero when idle.

Hobby is free with capped usage. Pro is $20 a month per developer seat with $20 of included usage, so a team of four pays $80 a month before usage. Usage is billed on active CPU time in milliseconds, provisioned memory for the life of each function instance, and the number of invocations.
For supported frameworks, deploying is connecting a repository and pushing. Vercel includes observability, automated preview environments for every pull request, instant rollbacks, infrastructure as code, and a CLI.



A Vercel project maps to one deployed application, so several services means several projects, much like separate Heroku apps.
There's no first-party Postgres. Vercel Postgres is no longer available, and existing databases moved to Neon in December 2024. New projects install a Postgres provider from the Vercel Marketplace, such as Neon, Supabase, Prisma Postgres, or AWS Aurora Postgres. Marketplace billing is integrated, like Heroku add-ons, but you manage each service in its provider's own dashboard.

Moving a backend from Heroku means refactoring web dynos into serverless functions and moving workers somewhere else. Frontends usually move easily. Functions have limits that long-running servers don't:
- Memory: up to 4 GB per function
- Execution time: up to 800 seconds (about 13.3 minutes)
- Size: up to 250 MB after gzip compression
- Cold starts: a new function instance adds latency on its first request. Bytecode caching and other optimizations make them less frequent but don't remove them
Those limits rule Vercel functions out for long-running work you may run on Heroku today: ETL jobs and large imports, video transcoding and image processing, large PDF and report generation, backups and provisioning tasks, invoice runs and payment retries, and bulk account operations. Work that holds a connection open, such as live chat, live dashboards, collaborative editing, delivery tracking, push notifications, and call signaling, doesn't fit either.
Northflank: a platform that can deploy into your own cloud account
Northflank suits teams that need workloads to run inside their own AWS, GCP, or Azure account, or that want a Kubernetes-backed platform without writing Kubernetes manifests.
Northflank runs on Kubernetes. You can deploy to Northflank's managed cloud, or use Bring Your Own Cloud, where Northflank provisions a managed cluster (such as EKS, GKE, or AKS) in your cloud account and deploys your workloads there. You can also import an existing cluster (BYOK). With BYOC, your data stays in your VPC and your existing cloud credits and commitments apply. You also own the Kubernetes cluster and the cloud bill, including node sizing, quotas, and cloud-provider configuration.
Northflank's pricing is based on the resources you provision, billed per second, with no seat fees. The free Sandbox is always on and includes 2 services, 2 jobs, 1 database, and up to $20 a month of resources, without autoscaling or HA. Pay as you go on Northflank's cloud costs $0.01667 per vCPU-hour and $0.00833 per GB-hour, so 0.5 vCPU and 1 GB of memory comes to about $12 a month. With BYOC, Northflank charges $0.02083 per managed node vCPU-hour, and your cloud provider bills the infrastructure.
Northflank's Heroku migration guide is manual. A single-dyno app becomes one combined service. A multi-dyno app becomes a build service plus one deployment service per process type, with a secret group for shared variables. Heroku Scheduler jobs map to cron jobs.
The managed PostgreSQL addon is available on all plans, with high availability on paid plans. You import data by uploading a dump or giving it your Heroku database's connection string.
You deploy from Git or a container image through the UI, API, CLI, or templates. Northflank exposes more building blocks than Heroku or Railway, such as build services, deployment services, secret groups, and addons. That gives you more ways to arrange things and more concepts to learn.
Each Northflank project lives in one region, chosen at creation. Multi-region means a project per region, across 17 managed regions or any region in your own cloud, connected with multi-project networking. Preview blueprints create a full environment for each pull request or branch, with builds, services, and databases, and expire it automatically. Persistent volumes are available on every plan, though a single read/write volume limits a service to one instance.
How do you migrate a Heroku app to Railway?
Most apps move with a repository connection, a variables paste, and a database restore. These steps cover a typical web app with a Postgres database. Railway's Migrate from Heroku docs have the full reference.
- Create an account on Railway. New accounts get a one-time $5 trial credit, valid for 30 days, with no credit card required.
- Choose Deploy from GitHub repo, connect your GitHub account, and select the repository you want to deploy. If your Heroku app runs
webandworkerdynos, deploy the repository once per process and set each service's start command.

- Copy your config vars. Export them from Heroku in
.envformat withheroku config --shell --app <your-heroku-app>. Then click the deployed service, open the Variables tab, click Raw Editor, paste the output, and save. You can also add variables one at a time with New Variable.

- Move your Heroku Postgres data. Add a Postgres database to your project, then:
- Turn on Heroku maintenance mode so no writes happen during the move.
- Create and download a Heroku backup, or run
pg_dumpagainst your HerokuDATABASE_URL. - Restore it with
pg_restoreorpsqlusing the Railway database's public connection string,DATABASE_PUBLIC_URL. If it's missing, add Public Access under the database's Settings → Networking. For a lower-downtime approach, see Migrate a production database into Railway with minimal downtime. - Point your service at the Railway database by referencing its
DATABASE_URLvariable, and remove the HerokuDATABASE_URLyou imported in step 3 so it doesn't override the new one. - Turn on backups, and HA if you need it, before cutting over.
- Give the service a domain. Click the service on the project canvas, open Settings, and go to Networking. Generate a Railway domain under
.up.railway.app, or add a custom domain and follow the DNS steps. - Check the app on its new domain before you move traffic. Open a few pages, run a request that reads from and writes to the database, and confirm your worker picks up a job. Then point your production domain at Railway and turn off maintenance mode on Heroku.
For larger databases, the Railway team can help with the move.
How do the platforms handle compliance?
All six alternatives hold a SOC 2 Type II report. They differ most on HIPAA: some include a BAA, some charge for one, and DigitalOcean's App Platform and Managed Databases aren't HIPAA-eligible.
| Platform | Certifications | HIPAA |
|---|---|---|
| Railway | SOC 2 Type II, SOC 3 | BAA as an add-on above a monthly spend threshold |
| Render | SOC 2 Type II, ISO 27001; GDPR DPA available | BAA on Scale or Enterprise, with a 20% surcharge on usage in the HIPAA workspace |
| Fly.io | SOC 2 Type II; DPA available | Self-service BAA at no extra cost on paid plans |
| DigitalOcean | SOC 2 Type II, SOC 3; public GDPR DPA | BAA through sales, but App Platform and Managed Databases aren't HIPAA-eligible |
| Vercel | SOC 2 Type II, ISO 27001:2022, PCI DSS; GDPR DPA | BAA as a paid add-on on Pro, or included with Enterprise |
| Northflank | SOC 2 Type II, ISO 27001 | BAA under an Enterprise contract |
Fly.io doesn't hold ISO 27001 itself; its data centers do. DigitalOcean's ISO 27001 covers its data centers, not DigitalOcean itself. Check each provider's trust or security page for current certifications before you move regulated workloads. Northflank lists its certifications on its security page.
Which Heroku alternative should you choose?
Choose Railway when you want to keep Heroku's git-push workflow and run your web process, workers, and databases together in one project. Config vars paste straight in, services share variables, and each service is billed for the CPU and memory it uses. You also get volumes, native cron jobs, multi-region replicas, and SOC 2 and HIPAA support without running Kubernetes. Plan for one service per Procfile process, and turn on Postgres backups before you cut over.
Choose Render when fixed instance prices suit your budgeting and you want fully managed Postgres in one region. Choose Fly.io when you want to place machines in many regions yourself and are comfortable setting up previews and rollbacks. DigitalOcean App Platform fits teams already on DigitalOcean that don't need volumes, environments, or preview deployments. Choose Vercel for frontends and Next.js apps, and host long-running backends elsewhere. Choose Northflank when workloads must run in your own cloud account and you're prepared to own the Kubernetes cluster underneath.
If Heroku still meets your needs, you can stay while you plan. Existing apps keep running.
Deploy your Heroku app on Railway: connect the repository you already have, paste in your config vars, and check the app on its new domain before you switch production traffic.
Frequently asked questions
Is Heroku shutting down?
No. In February 2026, Heroku moved to a sustaining engineering model focused on stability, security, reliability, and support. Existing apps keep running and existing contracts are honored, but Enterprise Account contracts are no longer offered to new customers.
Can I keep my Procfile on Railway?
Yes, with one process type per service. If your Procfile defines web and worker, create a service for each and set its start command.
How do I move my Heroku Postgres database?
Put the Heroku app in maintenance mode, take a backup or run pg_dump, and restore it into Railway Postgres with pg_restore or psql using DATABASE_PUBLIC_URL. For a lower-downtime option, see Migrate a production database into Railway with minimal downtime.
Further reading
These cover the migration in more detail and where to get help along the way.
- Migrate from Heroku
- Migrate a production database into Railway with minimal downtime
- Railway Discord
- Railway Help Station
- Book a call with the Railway team for larger workloads or specific requirements