Use your Apache Ossie model in Claude, ChatGPT and Cursor

Your semantic model stops at one tool

A semantic model takes real work: what each table holds, which column is revenue, how your team counts an active customer. Then it lives in the tool it was written for. When someone asks Claude or ChatGPT about the same database, the AI starts again from bare column names, picks the wrong table and reports amounts in the wrong units.

Bring it in once, and every AI tool reads it

Import the .ossie.yaml file on a connection’s Context page. SaturnSQL matches each dataset and field to your tables and columns, and from then on Claude, ChatGPT, Cursor and Claude Code get the descriptions, synonyms, notes and definitions that fit each question through SaturnSQL’s MCP server. Nobody has to explain the data to the AI again, on any database SaturnSQL connects to.

See what changes before anything is saved

The import shows what it will change first: which notes, tables, columns and definitions, and which datasets match nothing in this connection. It only adds and replaces. What your team wrote that the file doesn’t mention stays as it is, and the whole import is one entry in the change log, with your name on it.

Keep the model in Ossie

What your team adds in SaturnSQL, including corrections the AI saved during a conversation, exports as Apache Ossie YAML on every plan. Take it back to your semantic layer or to any tool that reads Ossie. Nothing you write is locked in.

version: 0.3.0.dev0
name: prod_postgres
ai_context:
  instructions: Amounts are in cents. Use orders_v2, not orders.
datasets:
  - name: orders_v2
    description: One row per paid order
    fields:
      - name: total
        datatype: Integer
        description: Order total in cents, before VAT
        ai_context:
          synonyms: [revenue, amount]

AI sees only what it should

Context comes with Guard on the Pro plan. Datasets for tables a person can’t see are skipped on import, and hidden tables and masked columns stay out of everything the AI reads.

How Ossie maps to SaturnSQL Context

Common questions

What is Apache Ossie? Apache Ossie, formerly Open Semantic Interchange (OSI), is an open YAML format for semantic models, incubating at the Apache Software Foundation under the Apache-2.0 licence since July 2026. Snowflake is a founding member, and Cube, AtScale, Kyvos and Databricks have joined. Cube imports and exports it, and SaturnSQL Context imports and exports it.

Where do I get an Ossie file? Cube imports and exports Apache Ossie from its CLI v1.8.3. You can also write the YAML by hand, or start from SaturnSQL’s own export of a connection and edit it.

Which versions of the format does SaturnSQL read? The core spec 0.3.0.dev0, a single model with its datasets, and the older layout that lists models under semantic_model, of which the first model is read.

Does importing overwrite what my team already wrote? Only where the file says something. A description in the file replaces the one in SaturnSQL, synonyms are added to the existing ones, and definitions are created or replaced by name. Nothing the file leaves out is changed or removed, and you see the changes before they are saved.

Does SaturnSQL run Ossie metrics like a semantic layer? No. SaturnSQL Context is not a metrics engine. A metric becomes a definition that tells the AI what the term means and which SQL computes it, and the AI still writes the query. If a number must match to the cent in every tool, keep that metric in your semantic layer.

Which plan do I need? Importing comes with SaturnSQL Context on the Pro plan, €29 per user per month, together with Guard. Exporting to Ossie works on every plan. Pro is coming soon. Join the waitlist to hear when it opens.

Can I import over the API? Yes. Send the file as yaml to POST /api/context/{connectionId}/import. Without apply: true it returns the preview and saves nothing. An API key needs the write scope, and the person behind it must own the connection or be an admin.

More on what AI tools get from your team’s notes is on the SaturnSQL Context page. Choosing a semantic layer? See the best semantic layer tools, compared on MCP and Ossie support.