At a glance
| SQL support | Rich analytical SQL; DuckDB 2.0 adds triggers, data-modifying CTEs and NEAREST joins |
| Storage and compute architecture | In-process, single-file columnar database; one read-write process, or several read-only processes; Quack for client-server access |
| Workload types | Local and embedded analytics, data science, ETL steps, file and lake querying |
| Integrations | Parquet, CSV, JSON, Iceberg, Delta Lake, DuckLake, S3, PostgreSQL, MySQL |
| Deployment | Library or CLI on Windows, macOS and Linux; WebAssembly in the browser |
| Licensing | MIT License, no enterprise edition; commercial support from DuckLabs |
DuckDB suits analysts, data scientists and engineers who want fast SQL over Parquet, CSV or JSON on one machine, in a notebook, a script or a pipeline step. It is not a shared multi-user warehouse by default, although DuckDB 2.0 adds a client-server option through the Quack protocol. It costs nothing to try.
Overview
DuckDB is an in-process analytical (OLAP) database: it runs inside a host program such as Python, R, Java, Node.js or its own CLI, stores data by column and executes queries with a vectorised engine. It reads Parquet, CSV and JSON files directly, on local disk or on object storage, and spills to disk for larger-than-memory operations. Its intellectual property is held by the non-profit DuckDB Foundation, and it is released under the MIT License.
DuckDB 2.0.0 LTS was released on 21 October 2026 and is supported until at least October 2027. It adds the Quack protocol, which lets a DuckDB process serve its databases to other clients, plus a VARIANT type, asynchronous I/O and a new PEG-based parser. In August 2026 DuckLabs, the company behind the core team, announced it would join AWS; the announcement states that the projects remain MIT-licensed under the DuckDB Foundation.
Key features
- In-process engine with no server to install, embedded in Python, R, Java, Node.js and other clients
- Direct SQL over Parquet, CSV and JSON files, locally or on S3-compatible object storage
- Extensions for Apache Iceberg, Delta Lake, DuckLake, PostgreSQL and MySQL
- Larger-than-memory processing that spills intermediate data to a temporary directory
- Quack remote protocol in 2.0 to serve databases to other DuckDB clients
- VARIANT type for semi-structured data stored in columnar form (2.0)
- Long-term support releases with a published release calendar
Pros and cons
Pros
- Free under the MIT License with no paid edition of the engine
- Nothing to provision: install a package and query files directly
- Reads open formats such as Parquet, Iceberg and Delta Lake
- LTS releases and a public release calendar help plan upgrades
Cons
- Limited to one machine's CPU, memory and disk
- Only one process can write to a database file unless you use Quack or DuckLake
- Database files written by 2.0 cannot be opened by earlier versions
Pricing
DuckDB is free and open source under the MIT License, and its FAQ states that there is no enterprise version. Your costs are your own hardware plus any cloud storage reads and egress. DuckLabs, the Amsterdam company that employs the core contributors, provides commercial support services; prices are not published. MotherDuck, a separate company, sells a hosted service built on DuckDB with a free Lite plan and a Business plan listed at USD 250 per organisation per month plus usage.
Pricing checked on the vendor's official page on 7 November 2026. Prices change; confirm before buying.
Who it is best for
Analysts, data scientists and engineers who want fast analytical SQL over files and lake tables on a laptop, a server or inside a pipeline, without running a warehouse.
Who should look elsewhere
- Organisations that need a governed, shared warehouse for many concurrent users: see DuckDB vs Snowflake.
- Transactional applications with many concurrent writers: see DuckDB vs PostgreSQL.
Verdict
DuckDB is an easy default for single-machine analytics: it is free, embeds anywhere and reads open file and table formats directly. Its limits come from the same design, one machine and a single writer per file, though DuckDB 2.0 and DuckLake widen the options. The move of DuckLabs to AWS leaves licence and governance unchanged according to the project, but is worth watching.
Frequently asked questions
Is DuckDB free for commercial use?
Yes. DuckDB is released under the MIT License, which permits commercial use, and the DuckDB FAQ states that there is no enterprise version.
Is DuckDB owned by AWS?
No. DuckLabs, the company that employs the core contributors, joined AWS in September 2026, but the DuckDB Foundation, a non-profit, holds the project's intellectual property and the project states it remains MIT-licensed.
What is MotherDuck?
MotherDuck is a separate, venture-backed company that sells a hosted service built on DuckDB. It is not part of the DuckDB project; check its own pricing page for plans.
Sources
- DuckDB FAQ (licence, Foundation, DuckLabs, MotherDuck)
- DuckDB release calendar
- Announcing DuckDB 2.0.0
- DuckLabs to join AWS, projects to remain open source
- DuckDB concurrency documentation
Checked November 2026.
How we research reviews and what star ratings mean: our editorial method.