Lightdash: self-hosted BI that treats dashboards as code
Agentic BI. Analytics at the speed of code ⚡️
At a glance
- What is it?
- Lightdash puts metrics, charts and dashboards in files, drives them through Git and CI, and adds AI agents and Data Apps on top of the same context layer. Here is what the repository actually documents, and where it stops.
- Who is it for?
- Adopt Lightdash if your metrics already live in a dbt project or you are willing to maintain a Lightdash YAML context layer, and you want dashboards reviewed in pull requests rather than clicked together in a browser. Do not adopt it if you need a BI tool that a non-technical analyst can run without Git, or if you cannot operate PostgreSQL plus object storage for the self-hosted path.
- Can I use it commercially?
- Check first. The repository uses a licence we do not classify automatically, so read its LICENSE file before any commercial use.
- Is it still maintained?
- Yes. The repository received new commits within the last day.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- lightdash/lightdash6,231 ★ · +77
| Repository | 2026-09-23 | 2026-11-15 |
|---|---|---|
| lightdash/lightdash | 6.2k | 6.2k |
77 Stars in the last 53 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
What Lightdash is for, and who it is aimed at
Lightdash describes itself as an open-source Agentic BI platform for modern data teams. The problem it targets is specific: in most BI tools the definition of a metric lives inside the tool, in a chart configuration that nobody reviews, while the warehouse and the dbt project live somewhere else under version control. Lightdash moves the metric definitions back next to the code. The README says your metrics, charts and dashboards live as files, and that you build them with coding agents, preview changes from the CLI, validate in CI, and review analytics in pull requests.
The audience follows from that. This is for teams that already have a dbt project or are willing to write Lightdash YAML pointed at a warehouse, and who have engineers or analytics engineers comfortable with a terminal. The README also lists business users as a group that can ask questions in plain English, explore dashboards, or create data apps, but those users sit on top of a context layer that someone technical maintains. If nobody on the team wants to own that layer, Lightdash is the wrong shape of tool.
The context layer is the whole design
Everything in Lightdash hangs off one artifact the README calls the context layer. Metrics, dimensions, joins, descriptions, caching and access rules are defined once there, and that single definition then powers dashboards, AI agents, data apps, embedded analytics, SDKs and MCP. The README states you can use dbt projects or standalone Lightdash YAML pointed at your warehouse.
That is the architectural bet. A dashboard in Lightdash is not the source of truth for a number; it is a consumer of a metric that already exists in the context layer. The same applies to the AI agents: the README says agents answer from your context layer, not raw table guesses, and that they respect permissions, return inspectable queries, reuse verified answers, and improve through reviews and evaluations. Whether the agent output is good depends on how much of your business logic you actually encoded, which is the same dependency dbt users already know.
The stack is a TypeScript monorepo. The frontend is React, Mantine, Vite and TanStack Query; the backend is Node.js, Express, TSOA, Knex and PostgreSQL. Warehouse adapters are listed for BigQuery, Snowflake, Redshift, Databricks, Postgres, Trino and ClickHouse. The root package.json sets Node 24 as the engine floor, so local development is not a lightweight affair.
Installing Lightdash and running your first preview
There are three documented paths. The README calls Cloud the fastest and recommended route, hosted at lightdash.com/start, with managed upgrades and no infrastructure. Self-hosting runs on your own infrastructure with Docker or Kubernetes, and the README points production deployments at a production deployment checklist and at separate Helm charts in the lightdash/helm-charts repository. The third path is local development from source:
git clone https://github.com/lightdash/lightdash.git
cd lightdash
./scripts/install.shThe README directs contributors to the CONTRIBUTING guide for the full local setup, package scripts and development workflow. The repository also ships a docker-compose.yml that defines a lightdash service on the lightdash/lightdash:latest image alongside a PostgreSQL database and a MinIO service for object storage, with the Lightdash port defaulting to 8080. Note that the compose file expects several variables to be supplied, including PGPASSWORD, LIGHTDASH_SECRET and the licence variables, so it is not a zero-configuration start.
Once the CLI is available, the README gives the agent-oriented workflow as three commands:
lightdash install-skills
lightdash preview
lightdash validateThe first installs agent skills so a coding agent has context for safe analytics changes. The second previews the branch, and the third validates the project before anything lands. The README frames this as the same workflow a team already trusts: preview the branch, validate the project, review the pull request, merge. If you want to look before installing anything, the README links a live demo at demo.lightdash.com/login.
Where Lightdash stops being the right tool
The governance model is also the constraint. Because metrics and dashboards are files, changing a chart is a code change: a branch, a preview, a validation run, a pull request. For a team of engineers that is the point. For a solo analyst in a company with no Git workflow, it is overhead that buys nothing, and the README offers no described path for editing a dashboard in the browser and committing it back.
The self-hosted path carries real operational weight. The docker-compose.yml in the repository brings up PostgreSQL and MinIO in addition to the application, and the production guidance points to a deployment checklist rather than a single command. The README also states that enterprise features are available on self-hosted deployments with a licence key, and that you should speak to the Lightdash team to get one. It does not enumerate which features those are, so anyone sizing a self-hosted rollout has to check the docs rather than the README.
Licensing is the other gap. The root package.json declares the package licence as MIT, but the repository metadata reports the licence as NOASSERTION, and the README distinguishes an open-source core from commercial enterprise features without stating the licence of each part. That is a question for the LICENSE file and the docs, not for the README, and it is worth resolving before a procurement conversation rather than during one.
Lightdash compared with Metabase and Superset
The two comparisons people search for most are Metabase and Superset, and the difference is where the metric definition lives. Metabase is built around a point-and-click question builder: an analyst creates a question in the UI, saves it, and it becomes a card on a dashboard. Superset takes a similar model, with datasets and charts configured in the application, and adds a heavier Python stack to operate.
Lightdash inverts that. The metric is defined in dbt or in Lightdash YAML, and the dashboard consumes it. The practical consequence is that two dashboards cannot quietly disagree about what revenue means, because both resolve to the same definition in the context layer. The cost is that adding a metric requires a code change and a deploy of the project, and that the tool assumes someone owns that project. Metabase and Superset will let a single analyst get to a chart faster on day one; Lightdash is aimed at the case where chart definitions have already become a governance problem.
Maintenance, releases and upgrade cost
The repository is not archived, and the last push was on 2026-11-01. Releases are frequent and versioned: 3.128.3, 3.128.2 and 3.128.1 all landed on 2026-11-03 according to the release list, which means the version number moves quickly and a pinned deployment will fall behind fast. There is a CHANGELOG.md at the repository root and an examples/upgrade-automation/ directory, which suggests upgrade automation is at least considered a first-class concern, though the README does not document a rollback procedure.
For Cloud users, the README states the platform is always up to date, which removes the upgrade question entirely and moves it into a vendor relationship. For self-hosters, the cost is the usual one: track the changelog, test the migration against a copy of the PostgreSQL database, and expect the compose file and environment variables to change between versions. The README does not document a supported downgrade path, so treat an upgrade as one-way until you have confirmed otherwise in the docs.
Editorial conclusion
Adopt Lightdash if your metrics already live in a dbt project or you are willing to maintain a Lightdash YAML context layer, and you want dashboards reviewed in pull requests rather than clicked together in a browser. Do not adopt it if you need a BI tool that a non-technical analyst can run without Git, or if you cannot operate PostgreSQL plus object storage for the self-hosted path. Before committing, verify three things against the docs: the exact warehouse adapter for your stack, whether the features you need sit behind an enterprise licence key, and whether your deployment target is Docker or the Helm charts.
Frequently asked questions
Is Lightdash open source?
Yes. The README describes an open-source core that you can self-host and contribute to, with commercial enterprise features available on self-hosted deployments via a licence key. Note that the repository metadata reports the licence as NOASSERTION even though the root package.json declares MIT, so check the LICENSE file for the precise terms.
What does Lightdash do?
It is a business intelligence platform built around a context layer that defines metrics, dimensions, joins, permissions and caching once. That layer then powers dashboards, AI agents, data apps, embedded analytics, SDKs and MCP, with the definitions stored as files rather than inside the tool.
How do I use Lightdash?
The README recommends signing up for Lightdash Cloud, or self-hosting with Docker or Kubernetes. For a code-driven workflow, the CLI provides lightdash install-skills, lightdash preview and lightdash validate, so you can preview a branch and validate the project before opening a pull request.
Is Lightdash free?
The README does not state a price. It says the open-source core can be self-hosted, and that enterprise features on self-hosted deployments require a licence key obtained by speaking to the Lightdash team. Pricing for Cloud is not given in the README.
How does Lightdash compare with Metabase?
Metabase is built around questions created in its interface, so a metric lives in the tool. Lightdash keeps metrics, charts and dashboards as files in your repository, built from a dbt project or standalone Lightdash YAML, and reviews changes through previews, validation and pull requests.
What is Lightdash?
The README describes it as an open-source Agentic BI platform for modern data teams, where a governed context layer defines trusted metrics, joins, permissions, business logic and caching once, and then powers dashboards, AI agents, data apps, embedded analytics, SDKs and MCP.
Official sources
Where this project is recommended
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/lightdash-lightdash)Star history chart for your README
[](https://hysenlabs.com/projects/lightdash-lightdash)Was this useful?