Open Catalog Access

Open Catalog Access shares tables managed in Treasure AI Complete CDP with your cloud data warehouse without copying data. You publish tables in Apache Iceberg — an open table format — to an export database, and Snowflake, Databricks, or BigQuery queries them in place through a managed Iceberg catalog: no export jobs, no transfer delays, no second copy to govern. Use it when teams outside the platform — data science, BI, AI engineering — need governed, read-only access to customer data in the engines they already use.

Enterprise data is fragmented across systems and teams, and forcing all of it into one warehouse rarely works. Without governed sharing, the CDP becomes a bottleneck: every team that needs customer data requests a custom export, and each export creates another copy to schedule, secure, and reconcile. Open Catalog Access replaces per-team exports with catalog-based access — Complete CDP remains the system of record, and external engines read the same governed tables directly. In terms of the Customer Intelligence Loop, it extends the Understand stage beyond the platform: analytics and AI teams work on unified customer data without waiting for exports.

Each Treasure AI account gets a dedicated set of AWS resources: an S3 bucket that stores Iceberg table data, an AWS Glue database (the default export database) that holds table metadata, and a read-only IAM reader role that external warehouses assume to read the data. Data engineers publish tables by writing them from Data Workbench with Trino; the warehouse connects to the catalog by assuming the reader role and reads table data directly from S3.

External engines connect as readers only — all writes go through Data Workbench, so Complete CDP stays the system of record. Warehouses read each table's current state on every query, which means data freshness matches your publish schedule: refresh a table with a scheduled query and every connected engine sees the update on its next read.

Tables are stored in Apache Parquet as Iceberg format version 2, and only the Trino engine reads and writes them from Data Workbench. See Reading and Writing Iceberg Tables from Data Workbench for the naming convention and supported SQL commands.

The BigQuery integration is experimental: functionality and interfaces may change without notice, and it is not recommended for production use.

• Open standard. Any Iceberg-compatible engine can read published tables — access is not tied to a proprietary export format or a single vendor.
• Seamless warehouse integration Publishing a table is one Trino query in Data Workbench; the warehouse side needs no pipelines, and new tables appear in the federated catalog automatically.
• Freshness follows publication. Engines read the table's current state at query time, so there is no separate sync step to schedule or monitor.

• Cross-engine analytics — data science, marketing, and BI teams query the same governed customer tables from the engine each team already uses.
• AI model training — train models directly on published Complete CDP data instead of maintaining separate export pipelines.
• Customer 360 data products — expose unified customer profiles as a governed, read-only data product in a data mesh architecture.

Open Catalog Access controls access in three layers:

1. Reader role trust policy — an AWS IAM trust policy defines which external services may assume the read-only reader role, scoped by IAM principal and external ID. See Manage the Reader Role Trust Policy.
2. Database-level access control — admins grant platform users READ, WRITE, or FULL access per Iceberg catalog database; this governs queries from Data Workbench. See Database-Level Access Control for Iceberg Catalog.
3. Read-only by design — the reader role cannot write, so external engines cannot modify published tables.

Access control policies for warehouse users are applied in the consuming engine with its native controls — for example Unity Catalog grants in Databricks or roles in Snowflake.

Both connect Treasure AI to your cloud data warehouse; the difference is where data is managed and which direction it flows.

Open Catalog Access is available in all Treasure AI regions: US (us01), Europe (eu01), Japan (ap01), and Korea (ap02). Each region has its own management API endpoint — see Use Iceberg Catalog Management API.

No. The reader IAM role is read-only by design, so external engines cannot create, modify, or delete published tables. All writes go through Trino queries in Data Workbench, and Complete CDP remains the system of record.

Warehouses read each table's current state on every query, so freshness equals your publish schedule. To keep a published table current, refresh it with a scheduled query in Data Workbench — no warehouse-side sync is involved.

No. Once an integration is set up, new tables and databases in the Iceberg catalog appear in the warehouse's federated catalog automatically.

Open Catalog Access publishes tables as Apache Iceberg format version 2, with Apache Parquet as the underlying data file format. Iceberg version 3 is not supported at this time, and neither are other table formats such as Delta Lake or Apache Hudi.

No. Published table data stays in the Treasure AI-managed S3 bucket, and external engines read it in place through the Iceberg catalog. Nothing is exported or duplicated into the warehouse.