Angelo Saraceno
PaaS is not a synonym for "the cloud." PaaS is a specific contract: I push code, the platform handles the runtime, the database is a managed primitive on the side, and the buildpack or Dockerfile figures out the rest. Heroku invented that contract in 2007. Some platforms still carry it honestly. Some have stopped. Some are doing other things and getting called PaaS for marketing reasons.
If you have already decided you want a real PaaS and not somewhere to run a container, this is the post for you. We make Railway, so this is partisan; I will also tell you, out loud, where the answer isn't us.
One house rule: every claim in this post is sourced. If I can't back something up I cut it rather than handwave.
PaaS in 2026, defined
A PaaS is the contract Heroku formalized: managed runtimes, the buildpack or Dockerfile abstraction, an add-ons marketplace for databases and queues and observability, and a deploy that starts with git push and ends at a public URL. In 2026 the contract has stretched to include first-class managed Postgres / Redis / Mongo, private service-to-service networking, multi-region deploys, and, increasingly, agentic deploys driven by Claude Code, Cursor, or a Stripe Projects CLI-style workflow.
Three things are not PaaS, no matter what the marketing pages say.
Container services (Cloud Run, AWS App Runner, which is moving to maintenance mode — announced March 31, 2026; stops accepting new customers April 30, 2026; existing customers continue) run your container at a URL. They do not provision your database, they do not own the buildpack abstraction, and they leave the integration story to you. They are a piece of a PaaS, not a PaaS.
Vanilla IaaS (EC2, GCE, Azure VMs) is the substrate. You can build a PaaS on top, which is what most providers on this list have done.
Serverless function platforms with no long-lived process (Lambda, Functions) are a different runtime contract, not the dyno contract this post is about.
The platforms, ranked
At a glance:
1. Railway
Best for full-stack apps that need a real backend, a real database, and a real bill.
I run Railway, so the bias declaration is up front. Railway is the PaaS where the runtime, the managed Postgres / MySQL / Redis / Mongo, the private networking between services, the cron jobs, and the multi-region for stateless services all live on one surface and one bill. One-click high-availability Postgres on Patroni shipped this March. Agentic provisioning via the Stripe Projects CLI (its stripe add command provisions managed infrastructure and is built for agent-driven workflows) shipped a few weeks later. We are the platform betting the hardest on agent-driven deploys in this listicle.
Features: git-deploy, native Postgres / MySQL / Redis / Mongo, one-click HA Postgres (Patroni), private networking between services out of the box, MCP server and Claude Code-friendly DX, Stripe Projects CLI agentic provisioning, per-environment secrets and config, persistent volumes, multi-region stateless services.
Pricing: $5 Hobby with included usage credit, Pro at $20/seat with usage on top. Usage-based pricing has a real downside I'm not going to talk around: if you don't watch the dashboard, a viral weekend can grow your bill. I tell every team to set spend alerts on day one. The trade-off is that you don't pay for what you aren't using.
Best for teams shipping a full-stack product that needs a real backend and a real database, not a marketing site. Teams who want one platform instead of stitching Vercel-plus-Supabase-plus-Upstash-plus-Render together. Teams who want agent-driven deploys to work end to end.
Honest trade-offs: cross-region multi-master Postgres writes aren't on the menu yet. Enterprise procurement (dedicated capacity, the deepest end of compliance attestations) is younger than AWS or Heroku.
Compare: Railway vs Fly, Railway vs Render, Railway vs Vercel.
2. Heroku
Best for institutional Heroku gravity (with an asterisk).
Heroku invented this category. Every other entry on this list owes a design debt to Heroku's dyno, buildpack, and add-ons model. I wrote that argument out at length earlier this year. In February 2026, Salesforce moved Heroku into sustaining-engineering mode and laid off roughly a thousand people across Heroku, Agentforce, and adjacent teams (reported Feb 2026). That changed the meaning of "Heroku is still a reasonable answer."
There are still legitimate reasons to stay: procurement gravity, compliance review cycles measured in quarters, an existing add-on contract you can't unwind this fiscal year. A platform in sustaining mode does not change those constraints, it changes how long you have to plan the exit.
Features: buildpacks, the add-ons marketplace, Heroku Postgres, Redis, ten regions (Mumbai and Montreal joined in late 2025). The Fir runtime, Heroku's Kubernetes-based generation, exists, but I have not seen a single greenfield team pick it.
Pricing: per-dyno, per-add-on, Salesforce-flavored. The Eco tier ($5/mo, sleeps) replaced the free dyno. The cheapest production-ready setup is meaningfully more expensive than its peers.
Best for teams who already know Heroku, have legacy apps deployed there, and have no urgent reason or political ability to migrate.
Honest trade-offs: the platform is in sustaining mode as of February 2026. Each app runs in one region; cross-region failover means duplicating apps and putting a load balancer in front. Pricing rose while the feature set did not.
Compare: Top Heroku alternatives.
3. Render
Best for the predictable, monthly-bill PaaS.
Render is the PaaS for teams who want the contract without surprises. If you need a web service, a worker, a Postgres, and you want a flat invoice you can defend in a budget meeting, Render is, for most teams, exactly that.
Features: git-deploy, managed Postgres, Key Value (Redis-compatible), cron jobs, private networking, autoscaling on Standard tier and above, preview environments.
Pricing: instance-based tiers starting at $7/month for the first paid web service. Managed Postgres in the same neighborhood. No surprise usage spikes.
Best for small to mid-size apps that fit cleanly into "web service, worker, database." Teams who care more about a predictable bill than about scale-to-zero economics.
Honest trade-offs: free-tier services sleep aggressively (15-minute idle timeout). Multi-region is effectively one region per service, with Frankfurt as the only EU option. The platform is opinionated in ways that work great until you hit one of the edges.
Compare: Railway vs Render.
4. Northflank
Best for teams committed to Kubernetes or bring-your-own-cloud.
Northflank is the PaaS-shaped abstraction on top of Kubernetes. Their bring-your-own-cloud story is strong, and they have leaned into enterprise AI and BYOC AI sandboxes as their wedge over the last year.
Features: Kubernetes-native scheduling, BYOC into AWS / GCP / Azure / Oracle / CoreWeave / bare metal / on-prem, preview environments, native managed databases, GPU support, microVM isolation, audit logging and SSO.
Pricing: per-instance with no markup on your underlying cloud cost when you BYOC (you keep your committed-spend discounts); management fees layered on top. Sandbox is free with a credit card.
Best for teams who have decided Kubernetes is non-negotiable, teams running AI workloads that need GPU plus microVM isolation, or teams with a compliance reason to keep workloads in their own VPC.
Honest trade-offs: the Kubernetes mental model leaks through in a way that, for most teams, adds surface area. If you wanted Kubernetes you would run it; if you didn't, Northflank's choice points are usually one more decision than you needed to make. Their own listicle's critique of Railway leans on framing ("smaller ecosystem," "limited advanced features") that has not tracked reality for a couple of years. The fact-check is yours to do.
5. Fly.io
Best for PaaS-shaped global apps, if you have ops capacity.
Fly is PaaS-shaped at the edges and IaaS-shaped at the core. The contract is closer to "deploying VMs" than "deploying applications," but the multi-region primitive is real and no other platform on this list has built it the same way.
Features: native multi-region deploys, edge instances, Firecracker-based microVMs, managed Postgres with regional read replicas, fly-replay header routing.
Pricing: per-VM, per-second machine billing. Legible once you understand the model, less legible if you don't. GPUs were deprecated for new accounts in 2024; if you need GPUs in 2026, look at Cloud Run, Northflank, or specialized providers.
Best for latency-sensitive services. Teams that have measured user latency and decided it matters. Anyone running an LLM endpoint, a multiplayer server, or an edge cache that needs to be in twelve places.
Honest trade-offs: reliability is the dominant 2026 critique. The October 2024 fleet-wide orchestration outage cast a long shadow, and the platform has had a steady drumbeat of incidents through 2025 and into 2026; check their public infra log before you commit. Ops surface is also higher than peers.
Compare: Railway vs Fly.
6. Vercel
Best for frontend-flavored PaaS.
Vercel is the PaaS for the frontend half of the stack. If your product is mostly Next.js, that is the right answer here and you can move on.
Vercel did ship a real answer to the long-running-cost critique in 2025: Fluid Compute with Active CPU pricing, which only charges for active CPU on idle-heavy or streaming workloads, can cut function bills by 80%+ on the right traffic shape. The compute cost story is now competitive on I/O-bound workloads.
Features: frontend-optimized CDN, native Next.js integration, Edge Functions, image optimization, preview deployments, Fluid Compute with Active CPU pricing, Vercel Postgres (Neon-backed), Vercel KV.
Pricing: Hobby free; Pro $20/seat with included credit, plus usage-based add-ons. Bandwidth and per-seat costs are the more legitimate critique in 2026; "Vercel cost me $286 on a $20 plan" is a genre of post for a reason.
Best for frontend-heavy applications. Static and ISR-heavy Next.js sites. Teams already standardized on Vercel's tooling. AI streaming endpoints that fit Fluid Compute's shape.
Honest trade-offs: the backend story is still several primitives in one trench coat (serverless functions, Edge Functions, hobby-tier limits that catch people on day 90). For long-running stateful workloads, you are better served pairing Vercel with a real backend PaaS.
Compare: Railway vs Vercel.
7. DigitalOcean App Platform
Best for budget-conscious small apps.
DigitalOcean's PaaS layer on top of their VMs is the "I know what DigitalOcean is and the bill will be predictable" answer.
Features: git-deploy, autoscaling on dedicated instances, static sites free tier, integration with DO managed databases (Postgres / MySQL / Redis / Kafka, each separately billed).
Pricing: static sites are free for the first 3 apps. Container web services start at $5/mo for shared-CPU. Managed Postgres adds $7+/mo on top, separately billed.
Best for small apps with predictable load, teams who already use DigitalOcean for VMs, projects where budget legibility matters more than developer experience.
Honest trade-offs: the developer experience is years behind the modern PaaS bar. The platform feels like a product line rather than a product. Support depth for production issues is the recurring complaint I hear when teams call to evaluate alternatives.
8. Encore.cloud
Best for typed-framework PaaS commitments.
Encore is the PaaS for teams who want to write Go or TypeScript against a framework that compiles down to deploys, infrastructure-as-config, and observability for free. It is opinionated in a way that the rest of this list isn't: you write Encore-flavored code, and you get the platform behavior in return.
Features: framework-driven services in Go and TypeScript, automatic infrastructure provisioning (pub/sub, cron, managed databases), traces and logs built in, deploy to Encore Cloud or BYO AWS / GCP.
Pricing: free tier; Pro starts at $39/seat/mo. BYO cloud available on higher tiers.
Best for teams that are willing to adopt the Encore framework as a coupling cost, in exchange for which they get an unusual amount of platform behavior automatically.
Honest trade-offs: framework commitment is real. You are not picking a platform, you are picking a code shape. If you ever leave Encore, you are rewriting the part of your app that talked to the framework's primitives.
9. Dokku
Best for self-hosted Heroku-clone on your own server.
Dokku is the OG self-hosted Heroku clone. It runs on a single VM, speaks Heroku-style buildpacks, and gives you the dyno-and-add-ons contract on hardware you already own. It is the open-source heart of this category, and it has aged remarkably well.
Features: buildpack-driven deploys, plugins for Postgres / Redis / MariaDB / others, Let's Encrypt integration, zero-downtime deploys, a CLI that intentionally feels like Heroku's.
Pricing: free. Your only cost is the VM you run it on.
Best for engineers who want the PaaS contract on a single small VPS, hobby projects with strict cost ceilings, teams whose compliance posture requires self-hosting.
Honest trade-offs: you are operating the VM, the OS updates, the disk, the backups, and the upgrades. The PaaS abstractions are real, but the operating cost is yours. If your VM dies, your apps die.
Six PaaS-shaped decision questions
Do you want the platform to provision your database, or connect to one you brought?
Railway, Render, Heroku, Northflank, DigitalOcean: provision. Fly and Vercel: provision in some shapes, BYO in others. Dokku: BYO via plugins.
Do you want a buildpack-style contract, or to write your own Dockerfile?
Heroku, Render, Dokku, DigitalOcean lean buildpack-first. Railway, Northflank, Fly are Dockerfile-friendly with buildpack-style detection for common stacks. Vercel and Encore are framework-specific contracts.
Do you want a hosted PaaS, or a self-hosted one?
Hosted: everything except Dokku. Self-hosted: Dokku, plus Northflank in BYOC mode (one foot in each camp).
Do you want it to run inside your own AWS / GCP / Azure account (BYOC)?
Northflank is the clear answer. Encore offers it on higher tiers. The rest are hosted-only.
Do you want it to handle background workers and cron, not just web services?
Workers + cron + web: Railway, Render, Heroku, Northflank, Fly, Encore, Dokku. Web-primary (workers possible but second-class): Vercel, DigitalOcean.
Do you want it to handle agentic / Claude Code deploys?
Railway. We have leaned into this the hardest of anyone in this list. The Stripe Projects CLI, the MCP server, and the Claude Code-friendly DX are all explicit bets on this being the next contract.
What happens next
PaaS as a category was about hiding infrastructure from application code. The next contract is going to be about hiding deploys from human developers entirely; the agent will provision, deploy, and reconfigure on behalf of the team. That is why every line item on Railway's roadmap right now is about making sure Claude Code, Cursor, and Codex can drive the platform without a human in the loop.
If you are evaluating PaaS in 2026, the decision is which contract you want to be on for the next three years: the one that hides VMs from your code (the old PaaS), or the one that also hides deploys from your developers (the new one). Most platforms on this list have not picked a position on that yet. Some never will.
Pick the platform whose contract matches the work in front of you. If your current PaaS has stopped fitting, the migration tax is real and it gets worse every month you delay. I see this on calls every week.
And if your current "platform" is a folder of Terraform you wrote yourself on top of EC2, the right move is to give yourself the quarter back.
Happy shipping.
Angelo
Angelo Saraceno is a Solutions Engineer at Railway. Before Railway he was at Citrix, working inside Verizon and Lockheed environments, so he has seen what "enterprise IaaS" looks like after the slides come down. He writes about infrastructure, deployment, and the gap between how cloud is sold and how it runs in practice.