DBeaver's dbvr tool runs SQL on PostgreSQL without a GUI, straight in pipelines and containers
dbvr, a standalone CLI from DBeaver Corporation launched in March 2026 that gained a free edition with the 26.1 release, promises to bring DBeaver's connectivity to locked-down servers, containers, and CI/CD pipelines without requiring a graphical interface or interactive login.

- Reading
- 6 min
- Tables
- 1
- Code
- 2
dbvr, a standalone CLI from DBeaver Corporation launched in March 2026 that gained a free edition with the 26.1 release, promises to bring DBeaver's connectivity to locked-down servers, containers, and CI/CD pipelines without requiring a graphical interface or interactive login.
What dbvr is and why it matters to those running PostgreSQL in production
DBeaver Corporation announced dbvr on March 26, 2026, as a standalone CLI built on the same multi-database engine used by the company's desktop products. Unlike the graphical client, dbvr doesn't open any window: it creates named connections and runs SQL straight from the terminal, without interactive login. That's exactly the profile missing on restricted servers, containers, and deploy pipelines, where installing a graphical client simply isn't an option.
The commercial edition is sold as an annual per-machine subscription, with a 14-day trial. DBeaver's 26.1 release brought dbvr Community, a free version under the Apache 2.0 license. That changes the adoption math: a team can validate whether dbvr fits into a CI/CD pipeline without negotiating a license first.
Among the capabilities DBeaver documents are native authentication for AWS, GCP, and Azure, result export straight to S3, Google Cloud Storage, or Azure Blob, and operation as an MCP server (mcp start) so an AI assistant can query a datasource. That last feature deserves caution: native cloud authentication and AI-driven access are capabilities that widen a production database's exposure surface if they aren't fenced in by a restricted access role.
Creating the datasource: from zero to the first SELECT
In dbvr, a datasource is a saved, named connection: subsequent commands refer to it by name instead of repeating host, port, and credentials on every call. The post on Planet PostgreSQL demonstrated the full flow using the classic DVD Rental sample database, with the postgres-jdbc driver:
dbvr datasource create --driver=postgres-jdbc --name=dvdrental --host=localhost --port=5432 --user=stoker --password=Secr3t --database=dvdrentalThe command returns an internal connection identifier, and dbvr datasource list confirms it was saved, linking the name dvdrental to the postgres-jdbc driver. From there, the sql command runs a one-off query or a full .sql script against any saved datasource:
dbvr sql -ds=dvdrental "SELECT * FROM actor LIMIT 5;"The result comes out tabulated right in the terminal, with row count and execution time at the end. The meta commands round out the basic set: they list databases, schemas, and tables, and print DDL, which lets you inspect an entire PostgreSQL schema with no GUI at all, useful when the only access available is an SSH tunnel to a production host.
Where this gains ground: pipelines, containers, locked-down servers
dbvr's value proposition isn't competing with psql in an ordinary interactive session. It's running where automation already runs. DBeaver and the post on Planet PostgreSQL list concrete uses:
- Scheduled reports: a cron job runs a saved
.sqlscript against the production datasource and exports the result as CSV. - Cloud export: the
sqlcommand reads a script and writes the result straight to an S3, Google Cloud, or Azure URI, with no intermediate manual upload step. - Pipeline verification: a deploy step runs a validation query, such as a row count or schema check, and fails the build if the return doesn't match expectations.
- Backup and restore validation: verification scripts run against a freshly restored database to confirm schema integrity and data completeness.
- AI access: dbvr can come up as an MCP server (
mcp start) so an AI assistant can query a datasource; in production, this should only point to a read-only datasource.
None of these cases is new in itself. What changes is having a tool with native cloud authentication and direct export to storage, instead of tying psql together with shell scripts that do a separate upload.
What a DBA requires before releasing this into production
Automation that runs without constant human supervision needs guardrails beforehand, not afterward. dbvr's own documentation and the post on Planet PostgreSQL recommend practices that, in practice, amount to basic governance for access to critical data:
- Separate projects for production. Mixing test and production datasources in the same project is the kind of mistake that only shows up once it has already caused damage.
- Password out of the command line. A password passed as a CLI option ends up exposed in shell history and in the process list; dbvr has
secret-managercommands and a credential pre-configuration guide to avoid this. - Read-only role in PostgreSQL. Granting only
SELECTon the schemas the report needs protects the data regardless of how dbvr is configured. statement_timeouton that role. A poorly written query that scans an entire table without an index needs to be canceled automatically, not lock up the connection until someone notices.- Versioned scripts. Running a reviewed
.sqlfile from a repository, instead of an ad hoc inline query, is what makes automation auditable.
On the risk of mixing up environments, the post sums up with humor something every DBA has lived through:
Yes, this never happens to you until it happens to you.
Post on Planet PostgreSQL
It's worth noting that none of these protections are exclusive to dbvr: they're the same rules that apply to any automation touching production. The difference is that the tool exposes cloud authentication and direct export to external storage, which widens the number of places where a poorly guarded credential can leak.
dbvr vs. psql: where each one wins
Neither one wins at everything. For PostgreSQL-only work on a server already under your control, psql remains more mature and has meta-commands (\d, \dt), \copy, and ON_ERROR_STOP that dbvr, according to the available documentation, doesn't yet have a confirmed equivalent for.
| Criterion | dbvr | psql |
|---|---|---|
| Cost | Annual per-machine subscription; free Community edition (Apache 2.0) | Free, already comes with PostgreSQL |
| Cloud authentication | Native for AWS, GCP, and Azure | Generally requires a token generated by the provider's CLI |
| Result export | Writes directly to S3, Google Cloud, or Azure | Writes a local file; upload is a separate step |
| Saved connections | Named datasources, grouped into projects, plus secret-manager | .pgpass and service files |
| PostgreSQL-specific features | No confirmed parity with psql | Meta-commands, \copy, variables, ON_ERROR_STOP |
The practical choice: psql for PostgreSQL-only work on infrastructure you already control; dbvr when the operation spans multiple database engines, uses a credential managed by a cloud provider's IAM, or needs to export directly to external storage. Many teams will reasonably use both, each at the stage where it fits best.
What remains open
dbvr's public documentation still doesn't detail full parity with psql's meta-commands, nor performance benchmarks under real production load. Before swapping a well-established psql script for dbvr in a critical pipeline, it's worth running the same set of checks in a staging environment and measuring behavior under statement_timeout, data volume, and network failure, exactly the scenarios a CI environment tends to expose before production.
Translated from the Brazilian Portuguese original · Read the original
PostgreSQL 19 brings lock-free REPACK and more robust logical replication for production operators
Released on November 19, 2026, PostgreSQL 19 replaces VACUUM FULL and CLUSTER with a REPACK command that runs without blocking reads and writes, and reduces manual work in logical replication. For those running databases in production, the real gain lies in the operational details.





