DuckDB 2.0 Gives Developers More Ways to Connect, Query and Build

Major open-source feature release advances the Quack protocol, accelerates remote data access, and expands SQL, extensions and AI-agent capabilities.

AMSTERDAM, NETHERLANDS, October 21, 2026 /EINPresswire.com/ -- DuckDB was built to run inside applications, without requiring a separate database server. Today, DuckLabs announced DuckDB 2.0, a major feature release that preserves this in-process architecture while expanding how developers connect to and work with data. A substantially improved Quack protocol enables communication between DuckDB instances over a network, while asynchronous I/O accelerates operations involving remote storage.

Quack debuted in preview in May 2026 in response to users who wanted shared database access. DuckDB 2.0’s new CONNECT statement allows a client to direct queries to a remote instance and receive results. Developers can continue embedding DuckDB in applications or explore using a shared instance when networked access is useful.

The release also introduces an extensible SQL parser, a new C API for extensions, triggers that respond to data changes, VARIANT support for semi-structured data, and CLI Agent Mode for AI coding agents.

“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. “We decided that this wasn’t a reason to abandon what works. Instead, it was a reason to give people another option. With DuckDB 2.0, you can keep running it inside your application or connect to a shared instance when that makes more sense. The choice is yours.”

One DuckDB instance, shared across a team
DuckDB 2.0 includes a thoroughly revised version of Quack, an extension that lets DuckDB instances communicate over a network. One process can serve its databases while another connects using CONNECT, sends queries to the server and receives results.

This offers developers an additional deployment choice. DuckDB can still run inside an application with no separate server to manage, or it can connect to a shared instance for analytical workloads needing networked access. The in-process engine remains at the center of both approaches. The Quack protocol is still being refined as the team works toward full production readiness.

Faster access to data in cloud storage
DuckDB already supports accessing data in remote storage, but synchronous I/O limited how many storage operations it could handle concurrently. In DuckDB 2.0, asynchronous I/O allows more read and write operations to proceed in parallel, with the I/O layer scaling independently of query processing.

This can reduce time spent waiting for data and accelerate analytical queries involving large datasets in cloud object storage, including Amazon S3. The improvement is not specific to one cloud platform. It is particularly valuable when queries spend more time waiting for remote operations than performing computations locally.

The release also improves query execution more broadly. In preliminary TPC-H decision-support benchmark testing at scale factor 1,000, a DuckDB 2.0 alpha build achieved composite scores approximately 25% higher than DuckDB 1.5.5. This reflects engine-wide improvements, not an isolated benchmark of asynchronous I/O.

Less rebuilding for extension developers
Previously, extension authors generally needed to rebuild and republish extensions for each DuckDB release, even if their code had not changed. DuckDB 2.0 introduces a redesigned, versioned C API, which makes it easier to build extensions for new releases of DuckDB.

Organizations can also host and sign their own trusted extension repositories, gaining greater control over distribution. Teams can manage their own extensions and choose which signing keys they trust rather than relying only on DuckDB-hosted repositories. Together, these changes reduce compatibility work and make extensions easier to maintain.

More capabilities for everyday data work
Triggers let DuckDB execute actions automatically when data changes, such as recording updates in an audit table. This can reduce the application code needed to track changes to database records. The VARIANT type adds end-to-end support for semi-structured data, including records with differing structures, helping users analyze logs and other evolving datasets.

CLI Agent Mode recognizes supported AI coding-agent environments, including Claude Code and OpenAI Codex, and adapts DuckDB’s command-line output accordingly. It presents compact results, flags truncated output and reports machine-readable errors. In one published test, it reduced query-output tokens by 59%, although overall agent cost and runtime did not decrease materially.

An extensible SQL parser lets extensions introduce new SQL syntax. An updated database-file format and improvements to compression and query execution further expand the release’s capabilities.

Built in the open
DuckDB 2.0 builds on work dating back to early 2026 and, for some components, considerably longer. Quack appeared as a preview release in May, and the team invited community testing of 2.0 development builds in September.

DuckDB 2.0 is designated a long-term support (LTS) release, with support through at least October 2027. DuckDB 1.5 support has been extended to March 2027 to give users additional time to test and migrate.

The release introduces a new default database-file format and a small number of breaking changes. Users upgrading existing deployments, particularly those involving multiple DuckDB processes or third-party extensions, should review the migration guidance before updating.

DuckDB 2.0 is available at duckdb.org/install. For a detailed technical overview of the new features and improvements, see the DuckDB team’s launch blog, Announcing DuckDB 2.0.0 LTS “Cinnamon Teal”. Release notes, documentation and migration guidance are also available at duckdb.org/news.

About DuckLabs
DuckLabs is the Amsterdam-based company behind the team developing DuckDB, an open-source analytical database designed to make analysis fast, portable and accessible. DuckLabs also develops DuckLake, Quack and other DuckDB ecosystem components. Since September 2026, DuckLabs has operated as a subsidiary of Amazon Web Services. Learn more at ducklabs.com.

Tim Cox
ZingPR
tim@zingpr.com
Visit us on social media:
LinkedIn
Bluesky
X