One machine, every engine

# Your database, your cache and your queue on one machine, one bill.

A small product usually needs a database, a cache and something to pass messages, and usually ends up paying for each separately, with its own provider, its own bill and its own way of failing. A Lathe instance runs any mix of Postgres 17, Redis 8, CouchDB 3, NATS 2 and SeaweedFS 4 on one dedicated machine, each engine on its own TLS port with its own password, sharing the machine's memory on a budget you set. It is one price, one backup and one place to look. [Start Free - $0 ](https://app.lathe.live/login)[From $15 a month → ](https://lathe.live/pricing)

## How the engines share a machine. Any mix Pick the engines when you create the instance, or add and remove them later. An engine you stop keeps its data on disk until you start it again. A memory budget you set Each engine gets a slice of the plan's RAM. Move the slider and the engine restarts with the new limit in about a minute. Each engine has a floor below which it will not go. Separate doors Every engine listens on its own port over TLS with its own password. One credential leaking does not open the others. One backup On paid plans the daily backup images the whole machine, so the database, the cache and the queue are restored together, to the same moment. One bill The plan price covers the machine, whatever runs on it. Adding an engine adds nothing to the bill. When to split If one engine needs the whole machine, or two workloads must not share a failure, run two instances. That is a choice you can make later, not one the pricing forces on you.

## Good for

- A web app with Postgres for data and Redis for sessions and cache
- Background jobs on NATS beside the database they work on
- CouchDB for documents that sync to devices, next to Postgres for everything else
- A staging environment that matches production on one small machine

## Questions

**Do the engines compete for memory? **

Each engine has a budget out of the plan's RAM that you set, and a floor it will not go below. Postgres alone gets the whole budget; add Redis and you decide the split.

**Can I add an engine to a running instance? **

Yes, from the portal, the API or MCP. The engine starts within a minute and gets its own port and password. Removing one stops it and keeps its data.

**Are the engines isolated from each other? **

They share the machine and its memory budget; each has its own port, TLS and password. They are not isolated from each other's load. For that, run two instances.

**Which versions and licences? **

Postgres 17 under the PostgreSQL licence, Redis 8 under AGPL-3.0, CouchDB 3 under Apache-2.0 and NATS 2 under Apache-2.0, all unmodified upstream builds.

Every number on this page comes from the [pricing page ](https://lathe.live/pricing)'s facts. [Previous Postgres on a cheap VPS, or the same box run for you ](https://docs.lathe.live/guides/vps-vs-managed)[Next pgvector embeddings on a dedicated Postgres machine ](https://docs.lathe.live/guides/pgvector)

## Related

- [Your own machine ](https://docs.lathe.live/machine)

Learn what Lathe means by a dedicated virtual machine: one customer per VM, with its own kernel, root filesystem and local disk.
- [Redis to Postgres: cache, queue, or one machine ](https://docs.lathe.live/guides/redis-to-postgres)

Cache-aside, queues and data moves between Redis 8 and PostgreSQL 17, with real commands and the decision criteria for one machine or two.
- [Host a container beside its Postgres, flat price ](https://docs.lathe.live/guides/host-your-app)

Put the image of your app on the dedicated machine that runs its database: deploy from a push, your domain, TLS, nothing metered. From $15 a month.

[All guides ](https://docs.lathe.live/guides)

## Postgres 17, Redis 8, CouchDB 3, NATS 2 and SeaweedFS 4 on one machine, from $15 a month.

Live about 5 minutes on average after creation on Free or checkout on a paid plan. Paid plans: cancel any time. [Start Free - $0 ](https://app.lathe.live/login)