Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/duckdb/duckdb/llms.txt

Use this file to discover all available pages before exploring further.

​
What are DuckDB Extensions?

DuckDB extensions are libraries containing additional functionality separate from the main codebase. Extensions provide added capabilities that can/should not live in DuckDB’s core for various reasons, such as:
  • Keeping the core codebase smaller and less complex
  • Managing licensing requirements
  • Providing specialized functionality for specific use cases
  • Enabling community-contributed features

​
Extension Loading Methods

Extensions can be integrated into DuckDB in two ways:

​
Static Linking

Extensions can be statically linked into DuckDB’s executables (CLI, unittest binary, benchmark runner, etc.). Statically linked extensions are automatically available when using these binaries without requiring additional installation steps.

​
Dynamic Loading

DuckDB has an extension loading mechanism to dynamically load extension binaries at runtime. This allows you to:
  • Install extensions on-demand
  • Keep your DuckDB installation lean
  • Update extensions independently of the core database

​
Extension Types

DuckDB extensions are divided into different types based on where they live and who maintains them.

​
In-Tree Extensions

In-tree extensions live in the main DuckDB repository. These extensions are:
  • Considered fundamental to DuckDB
  • Tie deeply into DuckDB internals
  • Expected to be affected by core DuckDB changes
  • Maintained by the DuckDB team
The DuckDB team aims to keep the number of in-tree extensions to a minimum and strives to move extensions out-of-tree where possible. Examples of in-tree extensions:
  • parquet - Parquet file format support
  • json - JSON data type and functions
  • icu - International Components for Unicode (collation, timezones)
  • autocomplete - SQL autocomplete functionality
  • tpch - TPC-H benchmark data generator
  • tpcds - TPC-DS benchmark data generator

​
Out-of-Tree Extensions (OOTEs)

Out-of-tree extensions live in separate repositories outside the main DuckDB repository. Moving extensions out-of-tree helps:
  • Keep the core DuckDB codebase smaller
  • Manage licensing requirements separately
  • Enable independent development cycles

​
DuckDB Managed OOTEs

These extensions are:
  • Distributed through the main DuckDB CI
  • Signed using DuckDB’s signing key
  • Maintained by the DuckDB team
  • Automatically distributed with every DuckDB release
Examples:
  • sqlite_scanner - Read SQLite databases
  • postgres_scanner - Read PostgreSQL databases
You can find the current list of DuckDB managed OOTEs in .github/config/out_of_tree_extensions.cmake in the DuckDB repository.

​
External OOTEs

These extensions are:
  • Not tied to the DuckDB CI
  • Running CI/CD in their own repositories
  • Maintained by external contributors
  • May or may not be signed, depending on the maintainer
The extension maintainer is responsible for testing, distribution, and ensuring an up-to-date version is available.

​
Extension Architecture

All types of extensions are built the same way:
  1. Using DuckDB’s root CMakeLists.txt as the root CMake file
  2. Passing the extensions to build as parameters
  3. Configuring how each extension should be built (loadable binary, linked into binaries, etc.)

​
Next Steps

Core Extensions

Learn about the core extensions included with DuckDB

Loading Extensions

Install and load extensions at runtime

Building Extensions

Build custom extensions from source