DuckDB 2.0 lets instances talk over a network and adds async reads from cloud storage

People who adopted DuckDB because it needs no server kept asking for one, and version 2.0 gives them a way to have both. DuckLabs, the Amsterdam company behind the open-source analytical database, has released DuckDB 2.0 as a long-term support version, supported through at least October 2027. DuckDB 1.5 support has been extended to March 2027 to give users time to migrate.

The headline change is a reworked Quack protocol, an extension that lets DuckDB instances communicate over a network. One process serves its databases, and another connects with a new CONNECT statement, sends queries and gets results back. Quack first appeared in preview in May 2026, and DuckLabs says it is still being refined towards full production readiness. Developers can carry on embedding DuckDB in applications.

“We designed DuckDB so people could analyze data without having to run a database server. Then they kept asking us for a server,” said Hannes Mühleisen, co-creator of DuckDB and CEO and co-founder of DuckLabs.

Faster reads from cloud storage

Synchronous I/O limited how many storage operations DuckDB could run at once. In 2.0, asynchronous I/O lets more reads and writes proceed in parallel, which should cut waiting time on queries over large datasets in object storage such as Amazon S3. DuckLabs says the gain is not specific to one cloud. In preliminary TPC-H testing at scale factor 1,000, a 2.0 alpha build scored about 25% higher than DuckDB 1.5.5, a figure that reflects engine-wide changes rather than the I/O work alone.

Extensions, triggers and agents

A redesigned, versioned C API means extension authors no longer have to rebuild and republish for every DuckDB release, and organisations can host and sign their own extension repositories. The release also adds an extensible SQL parser, triggers that run when data changes, and a VARIANT type for semi-structured data such as logs.

CLI Agent Mode recognises supported AI coding-agent environments, including Claude Code and OpenAI Codex, and returns compact results, flags truncated output and reports machine-readable errors. In one published test it cut query-output tokens by 59%, although overall agent cost and runtime did not fall materially.

There is a new default database-file format and a small number of breaking changes, so anyone running several DuckDB processes or third-party extensions should read the migration guidance first. DuckLabs has operated as an Amazon Web Services subsidiary since September 2026. DuckDB 2.0 is available from duckdb.org/install.

To stay across the latest in cloud, AI and enterprise tech analysis from Compare the Cloud, subscribe to our weekly newsletter at https://www.comparethecloud.net/newsletter

More News