BI Tool Self-Hosting and Licensing

Self-hosting a BI tool and running it for free are not the same thing. The license determines what you’re actually allowed to do with the code, what features you get without paying, and what obligations you take on if you modify and distribute the software. Three licenses dominate the open-source BI space — MIT, AGPL, and proprietary — and they produce meaningfully different outcomes for teams that self-host.

Lightdash: MIT License

Lightdash is MIT-licensed. That’s the most permissive open-source license in common use: you can run, modify, distribute, and even sell software built on it without releasing your modifications. There are no copyleft obligations.

The practical implications for a self-hosted Lightdash deployment:

• No user limits on the self-hosted version
• No license fees
• No obligation to open-source your customizations
• You’re responsible for all infrastructure: PostgreSQL for application state, S3-compatible object storage for cached results, and a headless browser (Chromium) for scheduled deliveries and PDF exports

Setting up Lightdash with Docker Compose for a small team is genuinely straightforward:

# docker-compose.yml (simplified)
version: '3'
services:
  lightdash:
    image: lightdash/lightdash:latest
    environment:
      - PGHOST=postgres
      - PGPORT=5432
      - PGUSER=lightdash
      - PGPASSWORD=${POSTGRES_PASSWORD}
      - PGDATABASE=lightdash_db
      - SECRET=${LIGHTDASH_SECRET}
      - SITE_URL=https://lightdash.yourdomain.com
    depends_on:
      - postgres
  postgres:
    image: postgres:15
    environment:
      - POSTGRES_PASSWORD=${POSTGRES_PASSWORD}
      - POSTGRES_DB=lightdash_db

For production use with more than a handful of people, the community Helm chart handles Kubernetes deployment. The infrastructure overhead is real but manageable for teams that already run Kubernetes.

What self-hosted Lightdash doesn’t include

The MIT license gives you the code, but the business model gates several capabilities behind the Cloud Pro tier ($2,400/month):

The split is typical of open-core models: the core product is genuinely useful, the premium tier adds operational features and integrations that most production deployments eventually want. For a small team running a proof of concept — testing whether dbt-native BI works for your workflow — the free tier covers everything you need to evaluate the tool seriously.

Metabase: AGPL License

Metabase uses the AGPL (Affero General Public License, version 3). AGPL is a copyleft license with a network use clause: if you modify AGPL-licensed software and provide it as a service over a network (not just internally, but to external users), you must release your modifications under the same AGPL terms.

For most internal Metabase deployments, this doesn’t matter. Your analytics team uses the tool internally; no software is being “distributed” or “provided as a service” to external parties. The AGPL obligations don’t trigger.

Where it gets complicated: if you’re embedding Metabase in a customer-facing product — building a SaaS tool where your customers access Metabase-powered dashboards — you may be subject to AGPL’s copyleft requirements. That means releasing your modifications (which might include your product’s integration code) under AGPL. Most commercial teams in this situation either license Metabase’s Commercial Enterprise edition (which removes the AGPL obligation) or choose a different tool for embedded analytics.

If you’re deploying internally and have no plans to embed Metabase in a product you distribute or sell access to, the AGPL license is effectively invisible in day-to-day operations.

What self-hosted Metabase includes

Metabase’s self-hosted offering is more feature-complete at the free tier than Lightdash’s:

• Visual query builder for non-technical users (full)
• Dashboards and charts (full)
• Embedding (basic, without white-labeling)
• SQL editor
• Alerts on query results

The governance-oriented features gate behind paid tiers:

If your organization needs SOC 2-ready audit trails or single sign-on, budget for at least Metabase Pro before running your internal cost comparison. The sticker price of “free” for open source understates the real cost for compliance-sensitive deployments.

Looker: Proprietary License

Looker has no free self-hosted option. The “original” Looker (customer-hosted) requires a paid license. Looker (Google Cloud core) runs only on GCP.

This isn’t just about cost — it reflects a different product philosophy. Looker’s value proposition is its enterprise governance model: the Permission Sets + Model Sets + Content Access permission system, detailed audit logging built in from the start, and LookML’s decade of refinement. That level of enterprise maturity comes with enterprise pricing.

The absence of a self-hosted free tier is a real filter. If you’re a startup or a small team evaluating BI tools, Looker effectively removes itself from consideration during the proof-of-concept phase unless your organization already has a contract. Teams that land on Looker typically do so because an enterprise procurement process determined it meets their governance requirements — not because an engineer tried it free one afternoon and loved it.

Infrastructure Costs Are Real

Even when the software is free, self-hosting isn’t. Both Lightdash and Metabase require:

• A server or container environment (EC2, Cloud Run, GKE, ECS — whatever your organization uses)
• A PostgreSQL database for application state
• Backups, monitoring, and maintenance

Metabase also requires a Java runtime (it’s a JVM application). Lightdash needs a headless browser for PDF/image exports of scheduled deliveries.

For a small team, running these on a single EC2 t3.medium or equivalent costs $30-60/month in infrastructure. That’s a rounding error compared to the $2,400/month Cloud Pro price point — but it’s not literally zero, and it comes with operational overhead that someone has to own.

Choosing Based on License

The license question simplifies once you know your constraints:

Use Lightdash self-hosted (MIT) if you’re a dbt team running a proof of concept, need unlimited users without per-seat cost anxiety, and are comfortable managing Kubernetes or Docker Compose infrastructure. Expect to upgrade to Cloud Pro when you need scheduling, caching, or AI features in production.

Use Metabase self-hosted (AGPL) if you need the richest free-tier feature set, have non-technical users who need a visual query builder today, and aren’t embedding Metabase in a distributed product. Check with legal before embedding in customer-facing surfaces.

Use Looker when the enterprise procurement process has determined it meets governance requirements. The lack of a free self-hosted option is intentional — the product is designed for enterprise procurement, not individual evaluation.

The BI Tool Selection Framework covers the broader trade-offs across these tools. For self-hosting specifically, the license determines your obligations and the feature tier determines what you actually get without paying.