Query · by Supercraft
Managed Hosting for DuckDB
Take the DuckDB database you already use locally and put it online. Query gives it dedicated CPU and RAM, NVMe storage, automatic backups and a remote Quack endpoint, so the database is shared, persistent and always on without you running the server.
- Native Quack
- Import your .duckdb
- EU and US regions
- Fixed monthly price
From $19 per month. No usage-based billing.
Plans Your DuckDB, remote What you get Portability How it compares Quack FAQ
-- from any DuckDB client, 1.5.3 or newer INSTALL quack; LOAD quack; ATTACH 'quack:mydb.query.supercraft.host' AS production; SELECT * FROM production.events LIMIT 10;
Your DuckDB. Remote.
Keep using DuckDB locally the way you do now. When you need shared, persistent compute, attach your Query database and run the statement against the remote server. Same SQL, same client, one more catalog in the list.
Quack is DuckDB's own client and server protocol, shipped in the core repository from DuckDB 1.5.3. It is in beta, and so are we.
Plans
Three sizes, one price each, billed monthly. What you pay for is the machine the database runs on: the CPU, the RAM, the NVMe and the backups. There is no per-query, per-scan or per-credit metering, because a database you cannot budget for is not one you can build on.
$19/mo
- CPU1 vCPU
- RAM4 GB
- NVMe25 GB
- BackupsDaily
- Quack endpointIncluded
A single database for a project, a dashboard or a CI job that needs a home.
$29/mo
- CPU2 vCPU
- RAM8 GB
- NVMe100 GB
- BackupsDaily
- Quack endpointIncluded
Room for a real analytical database and the people who share it.
$59/mo
- CPU4 vCPU
- RAM16 GB
- NVMe250 GB
- BackupsDaily
- Quack endpointIncluded
For the workloads that make a laptop fan spin: joins across Parquet, wide aggregates, long scans.
Configure your Standard database
Two choices, then deploy. Nothing is charged on this page.
Step 1 of 2: Region
Step 2 of 2: How do you want to start?
Dedicated instance, daily backups, Quack endpoint. Cancel monthly.
DuckDB without the server work
Remote access
Connect from your laptop, CI, an application or an analytics tool, without SSH tunnels or copying database files around.
Bring your existing database
Start from a .duckdb file you already have rather than rebuilding your data stack around a new engine.
Predictable resources
CPU, RAM and NVMe sized for your workload, on a machine that is yours. No mystery warehouse credits, no surprise at month end.
Managed durability
Automatic daily backups, monitoring and recovery, so nobody has to babysit another server.
Your database stays portable
Query does not turn your database into a proprietary warehouse. You import a DuckDB database, you work on a DuckDB database, and every daily backup is a DuckDB database you can download. If you want to leave, take the file and go.
The one thing we ask is a supported DuckDB version, because the storage format moves between releases and we keep the server current.
laptop
mydata.duckdb
|
v
+--------------------+
| QUERY |
| DuckDB + Quack |
| backups, monitor |
+--------------------+
|
v
backup.duckdb
(download any day)
When should you use Query?
| Local DuckDB | Query | Cloud warehouse | |
|---|---|---|---|
| Setup | Very easy | Easy | Varies |
| Always online | No | Yes | Yes |
| Remote access | No | Yes | Yes |
| Dedicated resources | Your machine | Yes | Usually abstracted |
| Existing .duckdb workflow | Native | Native | Usually not |
| Infrastructure to manage | None | None | None |
| Pricing | Free | Fixed monthly | Often usage-based |
If your data fits on your laptop and nobody else needs it, keep it local; DuckDB is excellent there and costs nothing. Query is for the moment that stops being true: a second person, a scheduled job, an app, or a dataset that outgrew the disk.
Native remote DuckDB with Quack
Quack is DuckDB's client and server protocol: a DuckDB instance becomes a server, and other DuckDB instances connect to it over HTTPS with DuckDB's own serialization instead of squeezing through a protocol built for a different database. Query is designed around it. You attach the remote database as a catalog, or run a one-off statement against it, from the same client you already have.
-- stateless, no ATTACH needed SELECT * FROM quack_query('quack:mydb.query.supercraft.host', 'SELECT count(*) FROM events');
Quack shipped as a beta in DuckDB 1.5.3 and its function names and defaults may still change; we track the releases and keep the endpoint matched to a current client. Connections are authenticated and TLS-only from outside the instance.
+-- S3 bucket
laptop -Quack-+
| QUERY
| DuckDB
+-- Parquet files
Put the compute next to the data
Keep large datasets in object storage and run DuckDB where the bandwidth is, instead of downloading them to a developer laptop for every query. Point the instance at your bucket, read Parquet in place, and keep the results in a table on NVMe.
S3 and S3-compatible storage at launch. Object storage is billed by your provider, not by us.
Questions
What is DuckDB hosting?
DuckDB normally runs inside your own process, on your own machine. Hosting it means running that same engine on a server that stays online, with your database on its disk, so several people, jobs and tools can use one copy remotely. Query is that server, managed: provisioned, backed up, monitored and updated for you.
Can DuckDB run as a server?
Since 1.5.3, yes: the Quack extension turns a DuckDB instance into a server that other DuckDB instances connect to over HTTP. Query runs that server for you with TLS, authentication, dedicated resources and backups.
What is the Quack protocol?
DuckDB's native client and server protocol. Requests and responses use DuckDB's own serialization, the client attaches the remote as a catalog with ATTACH 'quack:host' or queries it statelessly with quack_query(), and the default port is 9494. It is in beta as of DuckDB 1.5.3.
Can I upload an existing .duckdb database?
Yes, that is one of the three ways to start. Choose "Upload a .duckdb file" when you deploy. The file has to be readable by a current DuckDB release; older storage formats need one local EXPORT DATABASE and import first.
Can Query read S3 or Parquet?
Yes. Configure your bucket credentials on the instance and query Parquet, CSV or JSON in place, the same way DuckDB does locally with the httpfs extension.
Is my instance shared with other customers?
No. Each plan is a dedicated instance with its own CPU allocation, RAM, NVMe volume and network endpoint. Nobody else's queries run on it.
Can I connect from Python?
Yes. The Python package is a DuckDB client like any other: duckdb.connect(), then INSTALL quack; LOAD quack; and ATTACH the remote. The same applies to the CLI, R, Node and Java clients on a Quack-capable release.
Does Query work with dbt?
dbt's DuckDB adapter runs against a local DuckDB that can attach remote catalogs, so models can read from and write to your Query database through the attached catalog. A native adapter target is on the list once Quack leaves beta.
How is this different from running DuckDB locally?
Locally, the database lives and dies with your process and your laptop. On Query it is always on, reachable by more than one person or job, backed up every day, and running on resources you chose rather than whatever is left after your browser. The SQL is the same.
How is this different from MotherDuck?
MotherDuck is a serverless cloud service with its own pricing model and its own extension. Query is a dedicated machine running stock DuckDB with the Quack protocol, at a fixed monthly price, where the database file is yours to download at any time. Different trade-offs; pick the one that matches how you work.
Can I export my database?
Yes. Every daily backup is a standard DuckDB database file you can download, and you can run EXPORT DATABASE to Parquet or CSV whenever you like. Leaving is designed to be easy.
Where are Query servers located?
Europe (Frankfurt area), US East (Virginia area) and US West (Oregon area). You pick the region when you deploy; the instance and its backups stay there.
Query is an independent service by Supercraft. DuckDB is a trademark of the DuckDB Foundation. Query is not affiliated with, endorsed by, or an official product of the DuckDB Foundation.