Data in
safe waters

Data platform powered by DuckDB. Plain SQL transforms, Python ingest, single-file warehouse. Up and running in under a minute.

Get started →
The middle path

Most teams are stuck between spreadsheet chaos and enterprise overkill.

If you're running the business out of Excel and CSV exports, you've hit the ceiling. If you're paying Databricks/Snowflake plus four glue tools, you're buying cloud-scale capacity for laptop-scale data.

havn is built for the teams in between: real analytics, sensible price, self-hosted or Norwegian-hosted.

Option A
Excel and CSV exports
Free, feels free
  • Zero cost, everyone knows it
  • Someone opens the file every morning
  • No version control, no lineage
  • Breaks past a few million rows
  • Numbers disagree across copies
The middle
Real platform, sensible price
One tool, yours to run
  • DuckDB-fast analytics on your data
  • SQL transforms with DAG and PR review
  • Built-in CI/CD, quality checks, lineage
  • Scheduled runs, not someone clicking Refresh
  • One install, no glue tools to buy
Option B
Databricks/Snowflake + 4 glue tools
Metered, cloud-scale
  • Distributed compute you'll never use
  • Thousands a month for laptop-scale data
  • Separate bills for ingest, transform, catalog, reverse ETL
  • US-hosted, Schrems III exposure
  • CU math, warehouse sizing, surprise invoices

Three building blocks, no glue required

Engine

DuckDB under the hood

Embedded OLAP analytics in a single file. No servers to manage, no clusters to provision. Swap in DuckLake when you outgrow the laptop.

Transforms

Plain SQL, automatic DAG

Write standard SQL with dependency comments. havn resolves ordering, detects changes, and rebuilds only what's needed. No Jinja, no templating.

Pipeline

Python for ingest and export

Pull data from databases, APIs, and files with plain Python scripts. Push results wherever they need to go. Built-in connectors for the common sources.

Up and running in four commands

  1. Install
    One pip command. Python 3.10+ is the only requirement.
  2. Scaffold
    Creates ingest/, transform/, export/ with a project.yml config file.
  3. Run
    Ingest pulls data into landing, SQL transforms build through bronze, silver, gold.
  4. Explore
    Browse tables, run queries, and visualise the DAG in the web UI.
terminal 
# install
pip install havn

# scaffold
havn init my-project && cd my-project

# run pipeline
havn jobs run full-refresh
  bronze.raw_orders       done  0.2s
  silver.fct_orders       done  0.6s
  gold.revenue_daily      done  0.4s
  ✓ 3 models built in 1.2s

# open the web UI
havn serve
  • http://localhost:3000

One tool, not five

Databricks is compute. dbt Cloud is transforms. Fivetran is ingest. Atlan is catalog. Hightouch is reverse ETL. havn ships all of it in a single install, with the pieces the enterprise stack treats as upsells.

Not just smaller than the cloud stack, wider. PR review, CI/CD, lineage, quality checks, dashboards, notebooks, masking: built in, not bolted on.

havn web UI showing the project overview: pipeline health with running models, warehouse schemas, and live log output
Overview page, mid-run. Pipeline health, warehouse browser, live log, all in one view.
Ingest Python scripts + built-in connectors pull data into landing havn ingest
Transform SQL DAG with change detection, incremental builds, schema evolution havn transform
Explore Table browser, column stats, ad-hoc SQL editor, instant model preview havn query
Export Push results to APIs, databases, or files with plain Python havn export
Test & Quality Inline assertions, YAML contracts, anomaly detection on every run @assert
Lint SQLFluff integration catches style and syntax issues before they ship havn lint --fix
Git + CI/CD Full git workflow from the web UI, auto-generated GitHub Actions, data diff PR comments havn diff
Monitor Run history, freshness checks, schema sentinel, Slack alerts havn history
Dashboards Visual query builder, 25 chart types, filters, cross-filtering, multi-page layouts havn serve
Notebooks Interactive .dpnb notebooks with code sandboxing, inline outputs, pipeline integration havn serve
Connectors Postgres, MySQL, Databricks, REST APIs, CSV, Parquet. Pull data from anywhere havn connect
Macros Python functions auto-registered as DuckDB UDFs, callable directly in SQL @macro
Resources Per-category budgets, live capacity bar, cancellable tasks. See every query's cost havn resources
Storage Single-file DuckDB, or DuckLake with encrypted Parquet, time travel, multi-writer havn migrate
Every capability works two ways >_ CLI ⋄ Web UI

Two ways to run it

Same platform, same SQL, same UI. Pick where it lives.

havn Core

Self-host, free forever

Source-available under BSL 1.1. Runs on your laptop, your server, or your own cloud. Everything on this page, yours to run.

FreeOpen source Any machinePython 3.10+ Single fileDuckDB warehouse
Install locally →pip install havn
havn Cloud · Coming soon

Hosted in Norway, always awake

A managed instance on Norwegian infrastructure, on servers owned by a non-US company. Your queries hit a warm DuckDB in milliseconds, not a cluster booting up. Fixed monthly price, encrypted storage, team accounts, git-based deploys. Launching later this year.

NorwayNon-US infrastructure Always awakeNo cluster warm-up Fixed priceNo CU math
Follow on GitHub →for launch updates

What you're not signing up for

No vendor lock-in, no surprise invoices, no data shipped to a US cloud by default, no proprietary SQL dialect. The boring promises most data platforms quietly break.

  • Your data stays yours

    You choose where it runs: your laptop, your server, or a dedicated instance. No third-party queries, no analytics pixels, no training on your tables. GDPR-friendly by design.

  • Running in 60 seconds, not 60 minutes

    pip install, init, serve. That's it. No Terraform, no Kubernetes, no IAM roles. One file holds your entire warehouse.

  • Source-available, free to use

    BSL 1.1 licensed. Free for any internal or commercial use, source always visible, auto-converts to Apache 2.0 after four years. No usage tiers, no per-seat pricing, no "contact sales."

  • SQL you already know

    Standard DuckDB SQL. No proprietary dialect, no Spark translations, no vendor-specific functions. Your skills transfer in and out.

Own your data stack

havn is what happens when you take the best ideas from modern data platforms and make them simple enough to run anywhere. Source-available under BSL 1.1, free for your team, yours to run.

pip pip install havn
docker docker run -v $(pwd):/project -p 3000:3000 ghcr.io/chraltro/havn serve
Read the guide View on GitHub