A Heroku alternative when you want one machine and one flat price

Heroku is still one of the easiest ways to put an app online. You push code, it runs on a dyno, and add-ons give you a database, a cache and the rest. For a team that wants a polished platform and independent scaling, that is a good deal.

Lathe makes a narrower trade. The app, Postgres, sign-in and an optional cache run together on one machine that belongs to you, and you pick one monthly plan price. You give up Heroku's breadth to get a bill with fewer lines.

Prices below were checked against Heroku's official pricing page on October 4, 2026.

A Heroku stack is a sum of parts

On Heroku each piece has its own plan. The current pricing page lists dynos, Postgres and Key-Value Store separately.

The entry column is cheap. A Basic dyno, Essential-0 Postgres and a Mini Key-Value store add up to $15 a month. The Eco dyno is cheaper still, but Heroku lists it as sleeping after 30 minutes of inactivity and for personal accounts only.

The production-shaped column is where the sum grows. A Standard-1X dyno, Standard-0 Postgres and a Premium 0 Key-Value store come to $90 a month. These are not equal-sized machines, so do not read this as a performance match. It shows the shape of the invoice: every part has its own size and price, and a sign-in service or other add-on would be another line.

What Heroku gives you that one machine does not

Heroku's Standard dynos add horizontal scaling, preboot and zero-downtime deploys, and Performance dynos add autoscaling. Its Premium Postgres plans list high availability. Essential Postgres lists a maximum downtime of four hours a month. Standard and higher plans add rollback windows and follower databases.

If your product needs to scale web processes out, survive a failed database node on its own, or sit behind private networking, Heroku is built for that and Lathe is not.

What Lathe changes

Lathe plans are $15, $29, $49 and $89 a month. The price is for one machine. Apps, Postgres, Auth, Redis, CouchDB and NATS can share it, with storage, backups and 20 TB of outbound traffic included. There is no separate line for a second service, a database connection or an auth user.

Your app runs as a container on that machine, and the database connection strings arrive in its environment. Heroku apps that already run from a Dockerfile or a standard buildpack output move over without a rewrite. Apps that lean on Heroku-specific add-ons or the Procfile model need a small amount of rework.

The limits are real and stated up front. One machine means no high availability, no automatic failover, no horizontal scaling and no point-in-time recovery. Paid plans take a daily backup and keep seven. Support comes from two founders by email, answered within a business day.

Which one fits

Choose Heroku when you want the platform around the app: scaling a web tier, managed failover, a large add-on marketplace, or a team already built around it.

Consider Lathe when one small product fits on one machine, you want the app and its data close together, and you would rather choose a capacity tier than add up dynos and add-ons.

Try it on a staging copy

Move a staging app first, not your production system. Run the same workload on both and compare how it behaves and what each month costs. Compute your own total from the plans you actually use, because the right comparison is your stack, not a pricing page example.

Lathe offers a Free plan at $0 when the current plan list includes it, and paid plans start at $15 a month. Free has no daily backups, so keep real data on a paid plan.