MongoDB Atlas alternatives: flat pricing and a document database you can branch

Start with the part most posts like this bury. Layerbase does not host MongoDB. We cannot, for licensing reasons, so what we host is FerretDB: the MongoDB wire protocol implemented on PostgreSQL, Apache 2.0 licensed, built on the DocumentDB extensions Microsoft released under MIT and donated to the Linux Foundation. Your driver does not change. Your mongodb:// connection string does not change. Your find, insertOne, and ordinary aggregation pipelines do not change.

A few things do. Text indexes and the $text operator, change streams, and $lookup are the gaps that catch people, and the full compatibility rundown lists the rest. Check your codebase against that list before anything else in this post matters. If your app leans on those features, stop reading and stay on Atlas.

The prices

Atlas rates verified 2026-08-19 from mongodb.com/pricing, converted at 730 hours per month, compute only.

The Atlas monthly figures are compute alone. Storage past what the tier includes, backups, and egress all bill on top. Our prices are flat and include the storage listed, the backups, and the transfer.

At matching memory, Atlas lists at roughly 4x our flat price (M20 against our 4GB) and roughly 6x (M30 against our 8GB), before any of the metered extras.

One caveat that belongs right here rather than in a footnote: an Atlas M-tier is a three-node replica set. Our dedicated preset is a single node with backups and branching. That is a price-for-workload comparison, not an architecture-parity comparison. If you need a replica set today, Atlas has something we do not.

The performance question

We ran YCSB against MongoDB 8.0 and FerretDB 2.7 on the exact hardware behind the $65 preset, 4 vCPU and 8 GB of RAM, with a working set larger than RAM.

One hardware profile, a working set larger than RAM, and we did not benchmark Atlas itself. The full writeup carries every caveat, including the two misconfigurations the benchmark found in our own fleet and the point where the picture inverts. Read it before you quote these numbers at anyone.

Branching, which Atlas does not have at any tier

Atlas has snapshots. You can restore one into a new cluster, which is slow and costs you a whole cluster, so nobody does it per pull request.

Because a FerretDB database on Layerbase is a Postgres database wearing a Mongo interface, we branch it copy-on-write with the same mechanics we use for Postgres. Hit Branch, get an isolated copy of the data in seconds, point a preview deploy or a CI job or an agent at it, delete it when the pull request merges. No export, no restore, no Mongo-specific machinery. Branching with MongoDB shows the workflow.

This is the capability I would move for, more than the price. A document database you can fork per branch changes how you test migrations.

Moving the data

Paste a mongodb+srv:// connection string on the Atlas migration page. We connect once, copy every collection with its documents and indexes into a managed FerretDB database, verify the document counts against your source, and hand you a report naming anything that did not come across. Your Atlas cluster is read and never modified.

It is a one-shot copy rather than live replication, so plan a cutover window. The migration guide covers the compatibility checks worth running before you commit to a date.

When to stay on Atlas

Not a rhetorical section. Four real cases:

• You need a replica set or multi-region today. Our dedicated preset is a single node. Atlas is the answer until that changes.
• You are write-heavy at billion-document scale. The write picture inverts at that scale and our own benchmark post says so.
• You depend on Atlas Search, change streams, $lookup-heavy aggregation, or time-series collections. These are the FerretDB gaps, and working around them is a rewrite, not a config change.
• Your Atlas bill has never surprised you. If a Flex cluster at $30 covers you and nothing is on fire, flat pricing is not solving a problem you have.

The case for moving is strongest when you are paying M20 or M30 money for a working set that fits on one box, or when you have wanted a branch of your production data and settled for a restored dump.

Starting

FerretDB is not on our free tier. It starts on our lowest paid plan, and the pricing page has the current numbers for the shared plans and the dedicated presets.

If you already have data in Atlas, start at the migration page and paste the connection string. If you are starting fresh, create a FerretDB database at cloud.layerbase.com/create and point your existing driver at it.