Your databases and files
PostgreSQL, MySQL, SQL Server, Azure SQL, Oracle, Snowflake, BigQuery, Databricks, ClickHouse and Redshift, or CSV and Excel files queried through DuckDB.
Ask questions, generate reports or build them by hand, on data that stays on your servers.
Ask your data in plain words and get the answer with its chart.
Describe the report you need and the AI builds it, page by page.
Ask where a number comes from: the lineage is read from OpenMetadata.
PostgreSQL, MySQL, SQL Server, Azure SQL, Oracle, Snowflake, BigQuery, Databricks, ClickHouse and Redshift, or CSV and Excel files queried through DuckDB.
Joins and measures defined once,
used by every report.Row-level security:
each reader sees only their own rows.
Place and size each visual
where you want it, on several pages.With cross-filtering and drill-down.
Bar, line, combo, pie, scatter, table, pivot, gauge, treemap, waterfall, heatmap and map, plus your own custom visuals.
Plug in a local model (Ollama, LM Studio),
an OpenAI-compatible API, Anthropic
or your Claude subscription.You choose how much of the data
it may read.
Roles per workspace and public links.PDF and Excel exports, and reports sent by email
on a schedule (Cloud).Lineage from OpenMetadata.
The code is published under the GNU AGPL-3.0. You can read it, run it at work, change it, and nobody can take back a version you have installed.
Self-hosted, Open Report runs from your server against your own databases. Nothing is sent anywhere else, and there is no limit on users or reports.
Rather not run a server? The hosted version runs in France, under EU law.
Free when you host it yourself. Paid only if we host it for you.
Free
Self-hosted, forever
β¬0
Hosted in France
β¬20/month
Coming soon
On quote
Hosted in France
Yes. The self-hosted Community edition is free forever under the GNU AGPL-3.0: every feature, no user limit, no expiring trial. The hosted Cloud is a separate, optional service with a free tier and paid plans, for people who would rather not run a server.
You run it. The application sits on your infrastructure, queries your databases from there, and the data never travels to a vendor. Pricing is not per seat, so adding readers costs nothing. And the source is readable and forkable, which means the tool cannot be discontinued, re-priced or feature-gated out from under you. Point by point: Open Report vs Power BI, vs Looker Studio, vs Metabase.
PostgreSQL, MySQL, SQL Server, Azure SQL, Oracle, Snowflake, Google BigQuery, Databricks, ClickHouse and Redshift. CSV and Excel files can be imported directly and are queried through DuckDB.
Yes. Type a question in any language and the AI answers with the figures and the chart that shows them: ad-hoc analysis without writing SQL or building a report first. Ask it for a whole report and it generates the pages, which you can then edit by hand. More on AI data analysis.
The one you choose: a local model through Ollama or LM Studio, any OpenAI-compatible API, Anthropic, or your own Claude subscription. The administrator decides what the AI may read: the schema only, the schema and cached aggregates, or query results. With a local model, nothing leaves your servers. More on self-hosted AI.
Docker. One docker-compose.yml runs the published image (amd64 and arm64) as a single container: download it, generate three secrets, run docker compose up -d. A home server, a NAS or a Raspberry Pi is enough. You can also run it from source on a server with Node.js 22 or newer, with ./install.sh.
Not when you self-host: the SQL runs from your own server against your own databases, and nothing is sent anywhere else. If you choose the hosted Cloud instead, it runs on European infrastructure in France.
Yes. Row-level security filters the rows a user is allowed to read, at the semantic-model level, so it applies to every report built on that model. Workspaces and roles control who can view, edit or administer.
Yes. The AGPL-3.0 permits commercial use. The one obligation it adds: if you modify Open Report and offer the modified version to others over a network, you have to publish those modifications under the same licence. Using it unmodified, internally or with customers, requires nothing of you.
Three commands on any machine that runs Docker.
# Get the compose file
mkdir openreport && cd openreport
curl -fsSL -o docker-compose.yml \
https://raw.githubusercontent.com/cracky777/openreport/master/docker-compose.yml
# Generate the three secrets it needs
printf 'SESSION_SECRET=%s\nINTERNAL_TOKEN_SECRET=%s\nDATASOURCE_ENC_KEY=%s\n' \
$(openssl rand -hex 32) $(openssl rand -hex 32) $(openssl rand -hex 32) > .env
# Start it
docker compose up -d
Then open http://localhost:3001. The first account created becomes the admin.
One container for amd64 and arm64, so a home server, a NAS or a Raspberry Pi will do. Everything is kept in the openreport-data volume; upgrade with docker compose pull && docker compose up -d. On a public host, add COMPOSE_PROFILES=https to .env for nginx and a Let’s Encrypt certificate.
From source instead: clone the repository and run ./install.sh (Node.js 22 or newer), as the README explains.