A Snowflake-style analytics warehouse, open source and written in Rust — SQL over Parquet & Iceberg, dbt pipelines with lineage, published dashboards, and a full MCP control plane so your AI agents can run the platform, not just query it.
Use our hosted instance directly — the same build, running on scale-to-zero cloud infrastructure. Open the console, run SQL, browse pipelines and lineage, no setup.
Single-tenant by design — each organisation runs its own instance and connects its own LLM over MCP. Docker Compose for a single box, Helm for Kubernetes.
git clone https://github.com/EBD-Sweden/opensnow cd opensnow/deploy/demo # set your domain in .env, then: docker compose up -d
Rust + DataFusion over Parquet and Iceberg. Warehouses, materialized views, CTAS, external table registration — plus PostgreSQL wire compatibility, so dbt and BI tools connect natively.
dbt models run in dependency order on OpenSnow’s own engine, with a built-in scheduler, per-model SQL inspection, and a full lineage view from raw sources to marts.
Publish story-driven dashboards to Metabase via the API, or build native charts in the console with no code. Public links embed anywhere — like this one ↗.
A complete MCP control plane — 29 tools covering models, pipelines, schedules, warehouses and dashboards — over stdio or remote HTTP. Pluggable auth: your IdP’s OAuth 2.x / OIDC tokens, JWT, or static keys, with per-tool RBAC. Connect Claude or ChatGPT and manage the warehouse in plain language.
A full pipeline over public Eurostat & ECB data: 20+ dbt marts covering housing, inflation, household wealth, competitiveness and sovereign risk — queryable in the console, visualised in the “Krona’s Bargain” step-by-step dashboard ↗ and written up in our analysis of the weak krona.
EBD Sweden deploys OpenSnow for teams — hosted or in your own cloud — and builds the pipelines, dashboards and agent workflows on top.