What the Heck is Hardcharger?
Hardcharger is GizmoData’s open-source DuckDB 2.0 server, built on Warp SQL, an HTTP protocol streaming Apache Arrow. Every feature included; no license keys.
Hardcharger is a new open-source server for DuckDB 2.0, released on October 21, 2026, by GizmoData. It is written from scratch and licensed under Apache 2.0, making it a significant change from its predecessor, GizmoSQL. The primary reasons for creating Hardcharger include addressing issues with Arrow Flight SQL's transport, such as gRPC connections being terminated during reloads, and implementing built-in security features.
Chargewire, the new protocol used by Hardcharger, replaces Arrow Flight SQL and runs SQL over ordinary HTTP/1.1 and TLS. This design allows for shorter queries to complete in a single round trip and enables long queries to be submitted, get a handle, and poll for results later. Additionally, Chargewire supports big results being read as several partitions over separate connections simultaneously.
The security features of Hardcharger include bearer tokens that are not cookies, so browsers do not send credentials on their own, and there is no CSRF to deal with. Every handle is bound to the identity that created it and checked on every use, preventing unauthorized access. Furthermore, access control is checked within DuckDB against each statement's actual plan, ensuring that nested SQL, views, and macros cannot bypass permissions.
Hardcharger includes comprehensive features such as built-in security, role-based access control (RBAC) with users, roles, and grants down to the column level, row policies, column masking, tenants, per-catalog access rules, Open Policy Agent support, a tamper-evident audit log, statement instrumentation, query profiles, a Prometheus /metrics endpoint, a statement queue, timeouts, KILL SESSION, and a graceful shutdown.
Additionally, a web UI is included, compiled into the binary, providing a SQL editor, results grid, and schema browser that respects access control.
Hardcharger can be installed using either package or command-line options for various platforms. The author tested Hardcharger by loading a DuckDB file from Medium and making it available using the web interface. The process was relatively simple, making it easy to share specialized data collectors on a laptop with others.
Rewritten by urgent.news; may contain errors — check the original.