Provisa / Provisa and Starburst
Starburst centres on an engine. Provisa centres on a model.
Starburst is centred on SQL analytics in a managed Trino lakehouse, and governance, data products and agent access all build outward from that engine. Provisa is centred on a governed model that works over the engine you choose, Starburst included. Start by deciding which of those centres you need.
Where each fits
| You need to… | Starburst | Provisa |
|---|---|---|
| Run SQL analytics across many sources at scale | ✓ Managed Trino lakehouse; Gravity for tags and ABAC; cluster-scale execution | Works with Starburst as the query engine; adds GraphQL, Cypher and more protocols over the same data |
| Build an app API with transactional writes, governed | SQL analytics workload; writes are engine-mediated | ✓ Transactional writes routed direct to the source driver, sub-100 ms; governed on every protocol |
| Give AI agents governed access | MCP server (read-only SQL, public preview since January 2026) | ✓ MCP server where the OAuth token maps to a role; agents inherit the same six access-control layers as a person |
| Keep catalog, semantic layer and governance tools in sync | Gravity manages tags and ABAC; Alation, Atlan and a community Collibra utility crawl Starburst | ✓ Publishes to external catalogs; imports from your semantic layer; or run it standalone |
| Ask graph or path questions over relationships | SQL | ✓ Cypher and SPARQL as first-class query languages alongside SQL |
| Hand curated datasets to other teams | ✓ Data products (GA in Galaxy); Context Studio for glossary and business rules (private preview) | Governed domains and views; model is declarative config in git |
| Change engines later without rewriting queries or policy | The engine is the product | ✓ Swap to Trino, Starburst, DuckDB, ClickHouse, Postgres, Snowflake, Databricks or BigQuery |
| Run air-gapped or on a regulated network | Galaxy cloud or an enterprise cluster | ✓ pip install "provisa[embedded]": one wheel, no Docker, no JVM, no root |
Use them together
Register Starburst as the query engine in Provisa. Starburst keeps doing what it does well: SQL federation at scale, Gravity ABAC, and Galaxy data products. Provisa adds GraphQL, Cypher and SPARQL over the same data, governed transactional writes routed past the engine, and consistent policy on every protocol, MCP agents included. Starburst's data products and Gravity stay exactly where they are.
Your semantics and policy stop living inside the engine, so you can replace it later if you ever need to.
If you only need Starburst
Your consumers are SQL and BI tools. You want one vendor's managed Trino platform with governance (Gravity), curated data products, and agent access over SQL. You are committed to Trino as the engine. Starburst covers that well on its own.
If you only need Provisa
You need GraphQL, Cypher, SPARQL or transactional writes as well as SQL. Governance must hold identically on every protocol, agents included. You want to connect the model to your existing catalog and lineage tools instead of running everything inside one engine vendor. Or you need to run from a single wheel, on a laptop, without a cluster.