For a general BI tool that business users can pick up, start with Metabase (AGPL open source edition; paid plans for SSO and row-level permissions). For SQL-heavy analysts and a fully Apache-licensed tool, Apache Superset. For a light SQL-to-dashboard tool, Redash (BSD-2-Clause), now community maintained. If your metrics are defined in dbt, Lightdash. For reports written as code in SQL and Markdown, Evidence. For operational and time-series dashboards on PostgreSQL, MySQL or SQL Server, Grafana (AGPL-3.0).
"Open source BI" covers tools with very different licences. Some are permissively licensed with no paid edition of the code (Apache Superset, Apache 2.0); some are copyleft (Metabase's open source edition and Grafana, both AGPL); and some are "open core", where the core is open but enterprise features sit under a commercial licence in the same repository (Metabase, Lightdash). The licence matters if you plan to modify the tool, embed it in a product or offer it to customers, so each entry below states it exactly.
For head-to-head decisions see Metabase vs Apache Superset and Power BI vs Metabase. If you are weighing commercial platforms, see Power BI vs Tableau, Power BI vs Looker and Looker vs Looker Studio.
Quick picks
Graphical query builder plus a native SQL editor, and a single Java service to self-host.
Apache 2.0 with no paid edition of the project, SQL Lab and an embedded SDK.
Builds metrics and dashboards on top of your dbt project; MIT-licensed core.
Built-in PostgreSQL, MySQL and SQL Server data sources alongside metrics and logs.
How we chose
We considered BI and dashboard tools whose core code is published under an open source licence and that query SQL databases: Metabase, Apache Superset, Redash, Lightdash, Evidence and Grafana. To be included, a tool had to be released or actively committed to in 2026, have its licence stated in its official repository, and support self-hosting. For each we recorded the licence, the latest release where the repository shows one, self-hosting requirements as documented, how SQL users work in it, and what the paid edition or managed service adds.
Apache Kylin is not listed: it is an OLAP engine that pre-computes cubes for other BI tools to query, not a dashboard tool, so it belongs with data warehouses (see data warehouse vs database). Looker Studio (renamed Data Studio in April 2026) is free but not open source, so it is covered in Looker vs Looker Studio. The order below is not a ranking: general-purpose BI tools come first, then code-first tools, then Grafana.
At a glance
| Tool | Licence | Self-hosting needs | SQL access | Paid option | Main trade-off |
|---|---|---|---|---|---|
| Metabase | AGPL (open source edition); commercial licence for paid features | Java JAR or Docker; PostgreSQL or MySQL app database | Native query editor, SQL models | Metabase Cloud and self-hosted Starter, Pro, Enterprise | SSO and row-level permissions are paid |
| Apache Superset | Apache 2.0 | Web app, metadata DB, Redis, Celery; Kubernetes for production | SQL Lab; virtual datasets | Preset (third-party managed) | Most infrastructure to run |
| Redash | BSD-2-Clause | Docker Compose with PostgreSQL and Redis; 4 GB RAM minimum suggested | Query editor is the main interface | None (hosted service closed 2021) | Community maintained; about one or two releases a year |
| Lightdash | MIT core; separate licence for enterprise code | Kubernetes with Helm, PostgreSQL, S3-compatible storage | SQL in dbt models | Lightdash Cloud Pro, Enterprise | Needs a dbt project |
| Evidence | MIT | Build and host the app yourself (Docker or a static host) | SQL queries embedded in Markdown pages | Evidence Studio, Enterprise | Developer workflow, not drag and drop |
| Grafana | AGPL-3.0 | One server process (package or Docker) with its own database | SQL query editor per panel | Grafana Cloud, Grafana Enterprise | Built for monitoring, not ad hoc BI |
Metabase
Metabase's repository holds the open source edition, released under the AGPL, and commercial editions under the Metabase Commercial Software License. Business users build questions with a graphical query builder; analysts use the native query editor with variables, field filters, snippets and references to models and saved questions. Models are curated derived tables, built in the builder or in SQL. The latest release on GitHub when checked was 63.19 (1 October 2026).
Self-hosting is a single Java service: Metabase recommends Eclipse Temurin JRE 25 or the Docker image, and PostgreSQL or MySQL as the application database in production. It has 18 official database drivers, including PostgreSQL, MySQL, SQL Server, Oracle, Snowflake, BigQuery and ClickHouse. Paid plans (Starter, Pro, Enterprise; cloud or self-hosted) add SSO, row and column-level permissions and SSO-based embedding; prices are on Metabase's pricing page. See Metabase vs Apache Superset.
- SSO, row and column-level permissions and SSO embedding are paid features.
- AGPL obligations apply if you modify it and offer it to others over a network.
- The native editor runs single read statements only.
Apache Superset
Apache Superset is an Apache Software Foundation project under the Apache License 2.0, with no paid edition of the project itself. It offers a no-code chart builder, SQL Lab (a web SQL IDE), and datasets with virtual metrics and calculated columns as a lightweight semantic layer. Version 6.1.0 was released on 13 May 2026 and needs Python 3.10 or later.
It connects to any database with a SQLAlchemy dialect and Python driver; the documentation covers 81 databases, but drivers are not bundled and must be installed. Production deployment needs a PostgreSQL or MySQL metadata database, Redis, Celery workers and a beat scheduler, and the docs recommend Kubernetes. The embedded SDK uses guest tokens that can carry row-level security rules. Preset sells managed Superset with a free tier for up to 5 users.
- The most infrastructure to operate of the tools on this page.
- Database drivers must be added to your image or environment.
- Less approachable than Metabase for users who do not write SQL (editorial).
Redash
Redash is licensed under BSD-2-Clause. Its interface centres on the query editor: you write SQL (or the source's query language), visualise the result and add it to a dashboard. The repository says it supports more than 35 SQL and NoSQL data sources.
Status after the Databricks acquisition: Redash's end-of-life notice states that the hosted service at app.redash.io shut down on 30 November 2021 so the company could focus on SQL analytics inside Databricks, and recommended the open source version to self-host. The project continues on GitHub under community maintainers: v26.3.0 was released on 2 March 2026 and v26.9.0 on 24 September 2026, the latter fixing 13 security advisories. The previous releases were in January and August 2025, so expect a slow release cadence. Self-hosting uses Docker Compose with PostgreSQL and Redis; the setup guide suggests at least 4 GB of RAM, although its pre-built cloud images are documented as older versions.
- No vendor-backed hosted service or commercial support since 2021.
- Slow release cadence; apply security releases promptly.
- No graphical query builder or semantic layer for non-SQL users.
Lightdash
Lightdash builds BI on top of a dbt project: dimensions and metrics are defined alongside your dbt models, then explored and charted in the Lightdash UI. Its licence file places the code under MIT, except the packages/backend/src/ee directory, which has its own enterprise licence.
The self-hosting guide uses Kubernetes and Helm, a PostgreSQL database for Lightdash metadata (separate from your warehouse) and S3-compatible object storage; enterprise features need a licence key. Lightdash's pricing page lists the self-hosted open source option as free with core BI features and community support, Cloud Pro at USD 3,000 a month (October 2026), and custom-priced Enterprise.
- Assumes you use dbt; without it there is little to build on.
- Self-hosting requires Kubernetes, PostgreSQL and object storage.
- Several features (for example SSO, scheduled delivery and embedding) are listed on paid plans.
Evidence
Evidence describes itself as "an open-source, code-based alternative to drag-and-drop business intelligence tools". You write SQL queries inside Markdown pages and add chart components; the project builds them into a web app that can be versioned and reviewed like code. The repository is under the MIT licence.
You can self-host the built project, including with Docker or on platforms such as Vercel, or publish on Evidence Studio, the company's hosted environment with Git integration, branch previews and publish history. An Enterprise tier adds SSO, SCIM and audit logs. Pricing was not shown on the pages we checked.
- Requires comfort with SQL, Markdown and Git; no drag-and-drop editor.
- Less suited to ad hoc exploration by business users (editorial).
- Paid plan pricing is not published on the pages we checked.
Grafana
Grafana is licensed under AGPL-3.0-only, with some components under Apache 2.0 as described in its licensing file; the latest release when checked was 13.2.3 (29 September 2026). Microsoft SQL Server, MySQL and PostgreSQL are built-in core data sources, so you can write SQL in a panel and chart the result, alongside Prometheus, Loki and other sources.
It is a strong fit for dashboards over time-stamped data, such as order volumes per minute or job durations, and for database monitoring (see open source database monitoring tools). In our view it is less suited to ad hoc business analysis, because it has no semantic layer or business-user query builder comparable to Metabase or Superset.
- Designed around dashboards and time series rather than self-service BI (editorial).
- AGPL licence obligations if you modify and offer it to others.
- Some features are reserved for Grafana Enterprise or Grafana Cloud.
How to choose
Start with who will use it. If business users need to answer their own questions, Metabase is the most approachable; if analysts write SQL all day, Superset or Redash fit better. If your team already models data in dbt, Lightdash reuses that work instead of creating a second semantic layer. If reports should be reviewed and deployed like code, Evidence. If the job is operational dashboards on fast-changing data, Grafana.
Then check the licence against your plans: AGPL (Metabase's open source edition, Grafana) matters if you modify the code and offer it to others; Apache 2.0, BSD and MIT are permissive. Finally, be honest about operations. Metabase and Grafana are single services; Superset and Lightdash need several components and are easiest on Kubernetes; and every self-hosted tool needs upgrades, backups and security patching. If that is a burden, the managed options (Metabase Cloud, Preset, Lightdash Cloud, Evidence Studio, Grafana Cloud) are the trade-off.
Frequently asked questions
What is the best free open source alternative to Power BI?
For self-service BI with business users, Metabase's open source edition is the closest in spirit; for SQL-heavy teams, Apache Superset. Neither replicates Power BI's DAX semantic models or Excel integration. See Power BI vs Metabase.
Is Redash still maintained after Databricks bought it?
Yes, by the community. The hosted service closed on 30 November 2021, but the open source project continues on GitHub, with v26.3.0 in March 2026 and v26.9.0 in September 2026, which fixed 13 security advisories. Releases are infrequent.
Is Metabase really open source?
Its open source edition is, under the AGPL. Paid features such as SSO and row-level permissions are under the Metabase Commercial Software License and need a paid plan, whether self-hosted or on Metabase Cloud.
Which open source BI tool is easiest to self-host?
Metabase and Grafana run as single services (Metabase with a PostgreSQL or MySQL application database). Redash uses Docker Compose with PostgreSQL and Redis. Superset and Lightdash expect several components and document Kubernetes for production.
Is Grafana a BI tool?
Not primarily. Grafana is a dashboarding and observability tool, but its built-in PostgreSQL, MySQL and SQL Server data sources let you chart SQL query results, which works well for operational and time-series dashboards.
Sources
- Metabase pricing
- Metabase on GitHub (licences, releases)
- Metabase: connecting to databases
- Metabase: running the JAR file
- Apache Superset on GitHub
- Apache Superset on PyPI
- Superset: installation methods
- Superset: connecting to databases
- Preset pricing
- Redash on GitHub
- Redash releases
- Hosted Redash end of life
- Redash: setting up an instance
- Lightdash licence
- Lightdash: self-hosting
- Lightdash pricing
- Evidence on GitHub
- Evidence: deployment
- Grafana on GitHub
- Grafana data sources
Checked October 2026.
How we research these guides: our editorial method.