Skip to content

Pricing

Pay by the hour.
Keep the monthly ceiling.

Add credit, then run what you need. Applications and databases meter while they are running and cap at their advertised monthly price; AI inference meters per call, by token or by what the generation cost. No seats, no recurring subscription, and no transfer meter.

How credit moves

  1. 01

    Top up once

    PayPal credit, not a recurring subscription.

  2. 02

    Run a workload

    Apps and databases draw hourly; AI inference draws per call.

  3. 03

    Reach the ceiling

    A continuously running workload stops accruing after its monthly cap.

Apps

Pick a size

CPU and memory are enforced as real container limits, not a suggestion on a card.

Low-traffic sites, internal tools and staging.

CPU
1 vCPU
Memory
1 GB

$8 / month $0.0119 per hour

Get started

Room for a typical application runtime.

CPU
1 vCPU
Memory
2 GB

$16 / month $0.0238 per hour

Get started

A second core, for concurrent requests.

CPU
2 vCPU
Memory
2 GB

$24 / month $0.0357 per hour

Get started

The usual starting point for production.

Most popular
CPU
2 vCPU
Memory
4 GB

$32 / month $0.0476 per hour

Get started

Memory-hungry runtimes on two cores.

CPU
2 vCPU
Memory
8 GB

$48 / month $0.0714 per hour

Get started

Four cores, for sustained request volume.

CPU
4 vCPU
Memory
8 GB

$64 / month $0.0952 per hour

Get started

A large working set with room to spare.

CPU
4 vCPU
Memory
16 GB

$96 / month $0.1429 per hour

Get started

Eight cores, for CPU-bound work.

CPU
8 vCPU
Memory
16 GB

$128 / month $0.1905 per hour

Get started

The largest single-container tier.

CPU
8 vCPU
Memory
32 GB

$192 / month $0.2857 per hour

Get started

Prices in USD, excluding tax. Transfer is not metered on any plan — no allowance to exhaust, no overage to bill, no line item you did not expect.

Databases

Somewhere to put your data

PostgreSQL, MySQL, MariaDB, MongoDB and Redis, running beside your applications and private to their project by default. Priced off their own catalogue, because a database wants memory where a web process wants CPU — marking up the app tiers would misprice both.

Development, and production that stays small.

CPU
1 vCPU
Memory
1 GB
Storage
15 GB

$15.23 / month $0.0227 per hour

Get started

A real working set and more connections.

Most popular
CPU
1 vCPU
Memory
2 GB
Storage
34 GB

$30.51 / month $0.0454 per hour

Get started

Production databases doing steady work.

CPU
2 vCPU
Memory
4 GB
Storage
56 GB

$60.84 / month $0.0905 per hour

Get started

Sustained query load over a larger dataset.

CPU
4 vCPU
Memory
8 GB
Storage
120 GB

$121.80 / month $0.1812 per hour

Get started

Datasets that should stay in memory.

CPU
6 vCPU
Memory
16 GB
Storage
248 GB

$243.72 / month $0.3627 per hour

Get started

Large datasets and heavy concurrency.

CPU
8 vCPU
Memory
32 GB
Storage
504 GB

$487.56 / month $0.7255 per hour

Get started

The largest managed tier.

CPU
16 vCPU
Memory
64 GB
Storage
1,016 GB

$975.24 / month $1.4512 per hour

Get started

Storage is the disk each tier is sold with, not a quota the server applies. A database that outgrows its tier can keep writing, so size it conservatively and monitor disk usage. Every engine is available on every tier.

Backups, and their limits

Backups are handled for you

On or off per database, stored by Maxlayer in our own storage on the server your database runs on. No bucket to connect, no schedule to choose — and while a database's backups are on, it bills at 130% of its plan.

Redis is not covered

Redis is the one engine we have no way to back up, so treat a Redis instance as a cache rather than a store of record.

Restores stay in your hands

A backup is a dump file in Maxlayer's storage. There is no automatic restore action, so plan and test the recovery procedure your application needs — and export what you rely on before you need it at 3am.

Databases are private by default

Only your own applications can reach one. You can publish it on a port to connect a local client — that is a deliberate, confirmed action, and once open anyone who finds it can try to log in.

AI inference

Pay for what each call uses

Chat, embeddings, images and video draw from the same prepaid credit as your apps and databases. There is no separate AI subscription or payment method.

The model decides the basis

Most models carry a rate card: an input and an output price per million tokens, published before you call. Image and video models do not — the providers price those per image, per second or per megapixel. Nor do the few embedding models that take pictures and audio as well as text, which charge a different rate for each. Those are billed what the request cost to serve plus the same markup, and the exact amount lands on your usage page.

Models can be listed, repriced, or removed by the platform without a marketing-site deploy, so a table copied here would be stale exactly when you need it to be right.

Read the current catalogue

The public catalogue names the models available now, their capabilities, and what each one charges — token rates where there are token rates, and the provider’s own per-unit prices where there are not. It needs no account or key to inspect.

Browse live models and prices

No feature gates

Every plan includes everything

Nothing sits behind a higher price except more CPU and more memory. No seat counts, no application quotas, no transfer meters, no feature you have to upgrade to unlock.

  • Unlimited team members
  • Unlimited applications and databases
  • Unmetered transfer on every plan
  • All five database engines on every tier
  • Custom domains and certificate requests
  • Automatic deploys when a webhook is active
  • Build, runtime, and deployment logs
  • Environment variables and secrets
  • Scoped API keys
  • Role-based access control

Questions

Before you commit

How does billing actually work?

You add credit, and each running application draws from it every hour at its plan's rate. The smallest application tier costs $0.0119 an hour. Nothing is charged on a monthly cycle, there is no invoice date, and no card is stored.

Then why are prices shown per month?

Because that is the figure worth comparing, and because it is a ceiling. Hourly charges stop once something has paid its monthly price, which takes 28 days of continuous running — so the remaining days of a longer month are free, and no month costs more than the number on this page. Run something for part of a month and you pay for the hours it ran, nothing else.

What happens when my credit runs out?

Your apps are stopped, and your organization is put into read-only mode until you top up. New inference requests are declined too. Nothing is deleted — your configuration, domains, environment variables and deploy history are all still there. Adding credit lifts the suspension, and you choose which applications to start again.

Is there a free tier?

No. Every application runs in a real container on a real server, and we would rather charge eight dollars for that than ship a free tier we quietly throttle. Need something up for an afternoon? Four hours of the smallest plan costs about five cents.

Do you charge per seat?

No. Invite the whole team; it costs nothing. You pay for compute, not for who can log in.

Do you charge for bandwidth?

No. Transfer is unmetered on every plan — nothing to exhaust, nothing to overage, so the number you see is the number you pay. If that ever changes we will say so beforehand, not afterwards on an invoice.

Can I change an application's plan?

Yes, in either direction, from its settings. The new CPU and memory limits apply the next time the application deploys, and billing moves to the new hourly rate from that point.

How do I pay?

PayPal, in one-off top-ups. Nothing to cancel and no recurring authorisation — when you stop topping up, you stop paying. That is the entire cancellation flow.

What counts as an application?

One deployable thing: a web app, an API, a worker. A frontend and a backend is two apps on two plans — better you hear that here than at checkout.

How is AI inference priced?

It depends on the model, and the catalogue says which. Most bill by the input and output tokens they use, at a rate published before you call. Image and video models are not sold by the token at all — the providers price those per image, per second or per megapixel — and a handful of embedding models charge a different token rate for text, pictures and audio. Those are charged what the request cost to serve plus our markup, and the amount appears on your usage page like every other call. Everything draws from the same prepaid balance as your applications and databases, and the catalogue changes independently of this static page, so it is the place to check before you call.

Do databases cost extra?

Yes. A database is a container on a server exactly like an application, so it runs on its own plan and bills by the hour the same way. The smallest of each — a 1 GB app and a 1 GB database — is $8 and $15.23 a month, drawn from the same balance.

What happens if my database fills its storage?

Nothing stops it. The storage figure on a database tier is the disk it is sold with, not a quota the server enforces — we would rather tell you that than let you find out by assuming there is a safety net. We watch disk per server and will ask you to move up a tier before it becomes a problem for you or anyone else on that machine.

Do you back up my database?

We do. Backups are on or off per database and stored by Maxlayer in our own storage, so there is nothing for you to configure. While a database's backups are on, it costs an extra 30% of its plan. Two limits worth knowing: Redis cannot be backed up at all, and restoring is manual — a backup is a dump file, and restoring it means running it yourself.

Why do backups live with you rather than in my own bucket?

Because running storage is part of running the service. A bucket per customer was a configuration you had to make and a credential you had to guard, and a backup that silently stopped arriving only surfaced when you needed it. Here the toggle is the whole setup, and the surcharge is on the same prepaid balance as the database itself. One consequence we state plainly: if you leave, your dump files leave with us — export them first, because restore is manual and there is no self-serve download.

Can I connect to my database from my laptop?

Yes, but not by default. A database starts private to your own applications; publishing it on a port takes an explicit, confirmed action by an admin. We keep it off by default because a database on the public internet is a materially different risk, and we do not firewall the port for you — restricting who can reach it is yours to do on the server.

Need something this page does not cover? Ask us .

Start with the workload you have

Bring a repository or an image, choose the smallest sensible plan, and add capacity only when the work asks for it.

  • GitHub + GitLab
  • Container images
  • OpenAI-compatible AI
  • Pay as you run