Every managed platform publishes a low entry price and a long list of meters. To compare them with a self-hosted Dokploy server you have to pick a workload and price it end to end. Ours is deliberately ordinary: a Next.js web app, a Node.js API, one background worker, a PostgreSQL database and a Redis cache, for a team of two developers, running all month. The vendor figures below are from vercel.com/pricing, railway.com/pricing, render.com/pricing and heroku.com/pricing on 31 October 2026; where a page did not load its prices we say so rather than guess.
Two caveats before the numbers. Managed platforms bill some things by actual use, so an idle service costs less than the allocation we price; and they bundle things a VPS does not, chiefly a global edge network and zero server maintenance. We note both where they matter.
The stack and the server it needs
On a single server this stack needs about 4 vCPU and 8 GB of RAM: roughly 400 MB for Dokploy, 150 to 200 MB each for the web app, API and worker, 512 MB to 1 GB for PostgreSQL, 100 MB for Redis, and headroom for Docker builds. That is the Growth preset on the MassiveGRID Dokploy hosting page: 4 vCPU, 8 GB RAM, 128 GB NVMe, $19.16 a month on shared cores or $39.60 on dedicated cores, with 1 TB of traffic, one IPv4 and Dokploy installed free. We use the dedicated figure as the self-hosted bill, because a production stack deserves steady build times.
Vercel
From vercel.com/pricing: Hobby is $0, Pro is $20 per month per developer seat with a $20 usage credit included, unlimited viewer seats, 1 TB of fast data transfer included and $0.15 per GB after it, and function invocations from $0.60 per million. Pay-as-you-go beyond the credit is uncapped.
Our stack does not fit Vercel on its own: the API can run as serverless functions, but the worker, PostgreSQL and Redis need another provider (Vercel sells them through marketplace partners at the partner's price). Two seats are $40 before any usage. Add partner databases and a worker host and a realistic bill for this stack is $60 to $100 a month, rising with traffic because every gigabyte over 1 TB and every million invocations is metered. What you get for it that a VPS cannot match: the edge network, preview URLs on every pull request and no servers at all. Our post on leaving Vercel covers the migration path.
Railway
From railway.com/pricing: Hobby is $5 per month including $5 of usage, Pro is $20 per month per workspace including $20 of usage, seats unlimited. Usage is billed per second at $0.00000772 per vCPU-second and $0.00000386 per GB-second, which Railway itself rounds to about $20 per vCPU per month and $10 per GB of RAM per month, plus $0.05 per GB of egress.
Railway runs the whole stack, which makes it the cleanest comparison. At the allocation our stack needs, 4 vCPU and 8 GB, usage is about $80 + $80 = $160 a month if everything runs flat out; Railway bills actual consumption, so a mostly idle stack costs a fraction of that, perhaps $40 to $70, plus the $20 Pro fee. The honest range is $60 to $180 depending on load, with the top end reached exactly when your product is busy. Railway's advantages are real: deploy from Git with no server, scale by sliding a number, and pay nothing for idle capacity.
Render
Render's pricing page loads its tables dynamically and returned no figures to our fetch on 31 October 2026, so we quote only what Render's own documentation and support answers confirm: the smallest always-on web service, Starter, is $7 per month per service with 0.5 CPU and 512 MB of RAM, and Render renamed its instance types to "compute plans" on 26 August 2026. Third-party calculators list the next size (1 CPU, 2 GB) at $25 and a Pro workspace plan at $25 per month; treat those as unverified.
Our stack needs three always-on services (web, API, worker) plus managed PostgreSQL and Redis. Even at Starter sizes, which are too small for a production Next.js server, that is $21 before databases; at the reported $25 tier for the two web-facing services it is about $57 plus databases and the workspace plan, so roughly $80 to $120. Check render.com/pricing before relying on any Render figure; it changes often.
Heroku
From heroku.com/pricing: dynos are Eco $5, Basic $7 (512 MB), Standard-1X $25 (512 MB), Standard-2X $50 (1 GB), Performance-M $250 and Performance-L $500 per month; Heroku Postgres is Essential-0 $5 (1 GB), Essential-1 $9 (10 GB) and Standard-0 $50 (64 GB, 4 GB RAM).
For our stack: a Standard-2X for the Next.js app (1 GB is the minimum that renders comfortably), a Standard-1X for the API, a Basic dyno for the worker, Essential-1 Postgres and a Key-Value Store add-on (not priced on the page we fetched) come to $91 plus Redis, and Eco dynos are not an option because they sleep. Heroku's strengths remain the add-on marketplace and fifteen years of operational polish.
The table
| Platform | Monthly bill for the stack | Priced how | What is not included |
|---|---|---|---|
| Dokploy on MassiveGRID Growth (dedicated) | $39.60 ($19.16 on shared cores) | Flat per month, 1 TB traffic, discounts on longer terms | Your time to install (or a free ticket), no edge network |
| Vercel Pro, 2 seats | $40 before usage; $60 to $100 with databases and a worker elsewhere | Per seat plus metered transfer and invocations | Worker, Postgres and Redis (partner-priced) |
| Railway Pro | $60 to $180 by load | $20 plus per-second vCPU, RAM and egress | Nothing; all services fit |
| Render | About $80 to $120 (partly unverified) | Per service compute plan plus workspace plan | Databases priced separately |
| Heroku | $91 plus Redis | Per dyno plus add-ons | Redis add-on; Eco dynos sleep |
At this size Dokploy on one dedicated server is between 1.5 and 4.5 times cheaper than the managed platforms, and the gap widens with traffic, because the VPS price does not move until you add a tier of transfer ($1 for 2 TB, $9 for 10 TB). It narrows for a stack that idles most of the month on Railway, and it inverts for workloads that genuinely need a global edge.
What you give up, honestly
- Someone else's uptime for the platform. On a VPS the dashboard, Traefik and your apps share a machine. MassiveGRID answers the hardware half with Proxmox HA failover and Ceph storage under a 100% uptime SLA; the Dokploy half is yours to update, which is one command.
- Edge delivery. Vercel's network serves static assets from near the user. Put Cloudflare in front of your Dokploy domain if that matters.
- Usage-based elasticity. Railway charges nothing for idle services. A VPS is paid for whether busy or not; the price is also the ceiling.
- Add-on marketplaces. Dokploy's answer is its template catalogue (500+ open-source apps, including Redis, PostgreSQL, MinIO and Plausible) deployed next to your apps, which costs nothing beyond the server.
What you keep: Git push deployments with webhooks, preview deployments, automatic HTTPS, databases with scheduled backups, rollbacks, scheduled jobs and a web terminal, all in the free core.
Frequently Asked Questions
Is Dokploy cheaper than Vercel?
For anything beyond a hobby site, yes, by a wide margin. Vercel Pro is $20 per developer seat per month before metered transfer and invocations, and databases come from marketplace partners. Dokploy is free on a server you already pay for: $9.58 a month for the 2 vCPU, 4 GB MassiveGRID Starter preset, $39.60 for the dedicated 4 vCPU, 8 GB Growth preset that runs a whole stack.
Can Dokploy replace Railway or Heroku for background workers and databases?
Yes. Workers deploy as applications with a start command and no domain, databases (PostgreSQL, MySQL, MariaDB, MongoDB, Redis) are first-class services with scheduled backups to S3, and Compose stacks cover anything else. The difference is that everything shares one server you size yourself rather than being billed per dyno or per vCPU-second.
Where do these vendor prices come from?
From vercel.com/pricing, railway.com/pricing and heroku.com/pricing on 31 October 2026, and for Render from its documentation and support answers because its pricing page did not return figures that day; the Render $25 tier and $25 workspace plan are third-party reports and marked as unverified. Re-check the pages before quoting them; all four vendors have changed prices in the last two years.
The whole stack on one server: $39.60 a month
The Growth preset on the MassiveGRID Dokploy configurator, 4 vCPU, 8 GB RAM and 128 GB NVMe on dedicated cores, runs a web app, API, worker, PostgreSQL and Redis with headroom for builds, on a Proxmox HA cluster with a 100% uptime SLA and a 15-day money-back guarantee. Dokploy is free; we install it on request.
Configure a Dokploy server