Provisa / Works with
Keep your tools. Add the part that connects them.
Each of these products does its part well. The hard part sits between them: keeping meaning and access rules aligned across all of them. Provisa is a compiler and runtime for your data control plane. You declare the model once, and APIs, governance, catalog entries and semantic definitions are generated from that declaration and reach the tools you already run.
What each does, and what Provisa adds
| What you run | Examples | What it does well | What Provisa adds |
|---|---|---|---|
| Catalogs and governance | Atlan, Collibra, Alation, DataHub | Browse, search, classify and document data across every system. | Publishes every governed table, column, term and lineage edge to your catalog on each model change, so it shows what is actually enforced. |
| Engines and lakehouses | Starburst, Databricks, Snowflake, Microsoft Fabric | Run your queries at scale, with their own governance tools. | Registers each as an engine, alongside others or embedded DuckDB, and adds GraphQL, Cypher and more protocols over the same data under one set of rules. |
| Data virtualisation | Denodo | Queries sources live through virtual SQL views, with a mature optimizer and connector library. | Registers those views as governed tables over JDBC, and adds GraphQL, Cypher and governed agent access. |
| Semantic layers | dbt Semantic Layer, Cube, AtScale | Define metrics once for analytics. | Imports their models through Apache Ossie, each as a proposal you review, so the same definitions reach every protocol and agent under the same rules. |
| API layers | Hasura, PostgREST | Instant APIs over your databases, built for transactional workloads. | Adds SQL, Cypher and more protocols over one governed model, across all your sources. |
You stay in charge
Your meaning and your access rules live in the model, above any one product. So you can use the best engine for each job, change a choice you've outgrown, recover from one that didn't work out, or scale back a product that isn't earning its place. Nothing has to move on day one, and every one of those decisions stays yours.
Over time you may find you can do without some of them. A data catalog, dbt or Airflow can each become optional once the model is doing that job. If yours keeps earning its place, keep it.
You can also skip the platform-selection project and its months of meetings to pick the one engine everything must live in.
And the sources underneath
Connect any SQL database, any REST, GraphQL, gRPC or OData API, any file or object store, or your own code in Node, Python, Go or Rust, and query it in 5 minutes.
Every connector is built on one contract. A named source is a short configuration on a shared base, and it inherits that base's conformance tests. Each one has a page showing its base, what it pushes down, what it needs to connect and when it was last verified live. If yours isn't there, write it in your own language in about 15 minutes, and it gets the same catalog, governance, hosting and protocols.
Working with each
Catalogs and governance
Atlan, Collibra, Alation, DataHub. Provisa publishes to your catalog so it shows what is actually enforced.
Engines and lakehouses
Starburst, Databricks, Snowflake, Microsoft Fabric. Keep the one you have and register it as an engine in Provisa, alongside others or embedded DuckDB.
Data virtualisation
Keep Denodo for SQL virtualization. Provisa registers its views as governed tables and adds more ways in.
API layers
Hasura, PostgREST. Provisa adds SQL, Cypher, and more protocols over one governed model, across all your sources.
Semantic layers
dbt Semantic Layer, Cube, AtScale. Provisa imports models from these tools through Apache Ossie, the open standard for describing what metrics mean. Each import is a proposal you review before it registers. Use Provisa alongside your semantic layer, and scale it back over time if the model covers what you need.