Layerbase walkthrough video: migrations, branching, DuckDB and agents
I recorded a walkthrough of Layerbase that covers most of what the product does, in the order I would show it to someone sitting next to me. It runs a little over 20 minutes.
Watch the Layerbase walkthrough on YouTube
Every chapter below links to its spot in the video, and most of them also link to a written post that goes further than a demo can. If you only care about one part, jump to it.
Contents
- Engines and branching
- Moving off MongoDB Atlas
- A database branch for every Vercel preview
- MongoDB to PostgreSQL
- Leaving Supabase without losing your users
- One DuckDB query across four sources
- The CLI and the desktop app
- A coding agent you talk to from your phone
- Plans, hibernation and dedicated servers
Engines and branching
00:00 is the short version of what Layerbase is: one account for 19 managed database engines, where 17 of them can be branched. A branch is a full copy of a database you can break without touching the original, and it is the thread that runs through the rest of the video.
At 01:25 I show the part people tend not to expect: SQLite and DuckDB over the network. Both are embedded engines, a library that opens a file, so neither has a server you can connect to. Layerbase puts one in front of them, so a hosted SQLite or DuckDB database gets a connection string like any other. The written versions are SQLite file to cloud database and DuckDB from file to cloud.
Moving off MongoDB Atlas
02:24 is a live migration from MongoDB Atlas into DocumentDB, the MongoDB-compatible engine. You paste the Atlas connection string, the copy runs, and the report tells you what came across and what did not. Your application keeps its MongoDB driver.
The step-by-step version, including the Network Access setting that trips most people up, is Migrate from MongoDB Atlas to DocumentDB. If you are still deciding whether to leave, MongoDB Atlas alternatives covers prices and the cases where you should stay.
A database branch for every Vercel preview
04:55 connects GitHub and Vercel to a production database, a staging branch and preview branches. Every preview deployment gets its own database branch with the connection string injected, so two pull requests running migrations at the same time stop stepping on each other.
Written up in A database branch for every Vercel preview. For the MongoDB side of branching specifically, see Branching with MongoDB.
MongoDB to PostgreSQL
At 08:41 the same MongoDB data becomes PostgreSQL tables. The conversion proposes a schema and the relationships between collections, you review and edit that mapping, and only then does it copy. Verification runs before you switch anything over.
This is the one I would read before trying: Convert MongoDB to PostgreSQL with a mapping you can review.
Leaving Supabase without losing your users
10:43 migrates a Supabase project, including the auth schema. The rows in auth.users come across with their bcrypt password hashes untouched, so nobody has to reset a password, and they show up in the Auth console on the new database.
The thing to know is that Layerbase does not run Supabase's login service for you. Your users and their hashes are there, and you add a small sign-in route against them. Migrating from Supabase to Layerbase walks through both halves.
One DuckDB query across four sources
12:31 is my favorite part. A single DuckDB query joins a MariaDB table, a PostgreSQL table, a CSV file and a JSON file, with no export step in between. Then a quick look at DuckLake, which keeps analytical tables as Parquet files in object storage with a PostgreSQL catalog.
If DuckDB is new to you, start with What developers should know about DuckDB. The DuckLake details are in Introducing DuckLake on Layerbase.
The CLI and the desktop app
14:52 leaves the cloud and runs databases locally, first with the Layerbase CLI and then with Layerbase Desktop. Local databases branch too, which is handy when you want to try a destructive migration on your laptop before anyone else sees it. Branching any database covers the local workflow.
A coding agent you talk to from your phone
16:42 is Agent Servers. An Agent Server is a small Linux machine that stays on, with your repositories and your coding agent already set up. In the video I use voice mode on my phone to ask the agent for a code change, then have it open a pull request and merge it.
The longer explanation, including which agents work and what they cost, is in How to vibe code from your phone.
Plans, hibernation and dedicated servers
19:55 goes through the flat-rate plans, how an idle database hibernates and wakes on the next connection, the always-on setting for databases that should never sleep, and dedicated servers for when you want a machine to yourself. Current numbers live on the pricing page and the dedicated pricing page, since they change faster than a video can. Sleep vs pause vs archive explains the idle behavior in detail.
The wrap-up at 21:52 is short. If something in the video is not covered here or in the docs, that is a gap I want to know about: ops@layerbase.com.
You can follow along on the Free plan. Create a database and try whichever chapter caught your eye.
Keep reading
- Migrate from MongoDB Atlas to DocumentDB, then branch itA step-by-step move from MongoDB Atlas to DocumentDB on Layerbase: the Network Access setting to change first, what the copy does and skips, how to check the result, and the branch you get at the end.
- Migrating from Xata to LayerbaseXata is unmodified Postgres now, so leaving it is a dump and a restore. Here is what copies across, what stays behind (branches, pgroll history, anonymization rules), the connection-string detail that makes or breaks the copy, and how to rebuild your preview branches on the other side.
- The SQL You Already Know Can Help You Grow Your SaaS: A Developer's Guide to DuckDB 2 and DuckLakeDiscover customer trends, understand product usage, and find opportunities for growth using familiar SQL. Follow along for free on Layerbase.
- DuckDB 2.0 is now on LayerbaseDuckDB 2.0 is live on Layerbase Cloud. Run analytical SQL through a hosted connection, explore your data, and build reports. Query account PostgreSQL and MariaDB sources and save snapshots in managed DuckLake.