Zero-copy warehouse access
Agents query Snowflake, Databricks, and BigQuery in place through native sharing. No extract, no second copy, no drift.
Platform
Data Fabric. Agents read your data where it lives.
Zero-copy access to Snowflake, Databricks, and BigQuery, governed SQL, an Apache Iceberg lakehouse, and more than 3,000 prebuilt connectors. Row and column security is inherited, not reimplemented.
app.meridian.example/data
3,012 connectors
Sources
8 connected
Snowflake
Zero-copy
4 min agoread in place
Databricks
Zero-copy
9 min agoread in place
BigQuery
Zero-copy
2 min agoread in place
HRIS
Connector
12 min ago48,212 rows
Governed SQL
as d.okafor · Snowflake
select cost_center, sum(actual) - sum(plan) as variancefrom finance.gl_actuals ajoin finance.plan p using (cost_center, period)where period = '2026-Q3'group by 1order by 2 desclimit 5;| Cost center | Variance | ||
|---|---|---|---|
| Field operations | 4,120,000.00 | 4,388,412.55 | +268,412.55 |
| Cloud infrastructure | 1,860,000.00 | 2,004,910.12 | +144,910.12 |
| Contract labor | 940,000.00 | 1,038,220.00 | +98,220.00 |
| Travel | 310,000.00 | 356,114.40 | +46,114.40 |
| Marketing programs | 1,250,000.00 | 1,231,880.75 | −18,119.25 |
5 rows · 412 msrow policy: cost_center in user.cost_centersfreshness: 4 minlogged: READ-88213
What is the Data Fabric?
Agents are only as reliable as the data they read. Copying HR and finance data into a vendor's store creates a second security model, a freshness lag, and a compliance question. The Data Fabric avoids the copy.
Agents query your warehouse directly through zero-copy sharing with Snowflake, Databricks, and BigQuery. Row-level and column-level security defined in the warehouse apply to every agent query. For systems without a warehouse footprint, more than 3,000 prebuilt connectors bring HRIS, ATS, payroll, ERP, CLM, and collaboration data into an Apache Iceberg lakehouse that you own and that your own tools can query.
Every agent read is a SQL statement or an API call you can inspect. Freshness is measured per source and shown to the agent and the user, so a variance explanation states which day's actuals it used. Nothing in the Data Fabric is used to train models.
- Design partners connected a warehouse and the first two source systems in the first week
- Warehouse security policies apply unchanged; no Meridian-side permission model to maintain
- Iceberg tables are readable by Spark, Trino, DuckDB, and the warehouses themselves
- Data residency follows your warehouse region; the Fabric runs in the EU or US to match
Capabilities
What Data Fabric does
6 capabilities, each enforced below the agent so they apply to Meridian, partner, and Studio-built agents alike.
Inherited security
Row-level and column-level policies defined in the warehouse or source system apply to every agent query. Agents cannot see more than the user they act for.
Apache Iceberg lakehouse
An open-format lakehouse you own, for sources without a warehouse footprint. Query it with the engine of your choice.
3,000+ prebuilt connectors
HRIS, ATS, payroll, ERP, CLM, banking, expense, ticketing, and collaboration systems, with incremental sync and schema tracking.
Governed SQL
Every agent read is an inspectable SQL statement or API call, logged with the user context and the result row count.
Freshness contracts
Per-source freshness measured and surfaced to agents and users. An answer says which day's data it used.
Specification
Data Fabric specification
Standards, protocols, and limits as of September 2026. Your account team can confirm coverage for a specific stack.
| Item | Specification |
|---|---|
| Standards |
|
| Zero-copy sources | Snowflake, Databricks, and BigQuery through native sharing; no extract, no second copy |
| Lakehouse | Apache Iceberg tables you own; readable by Spark, Trino, DuckDB, and the warehouses themselves |
| Connectors | 3,000+ prebuilt connectors for HRIS, ATS, payroll, ERP, CLM, banking, expense, ticketing, and collaboration systems, with incremental sync and schema tracking |
| Security | Row-level and column-level policies from the warehouse or source system apply to every agent query |
| Query | ANSI SQL; every agent read logged with user context and result row count |
| Freshness | Measured per source and shown to agents and users; answers state which day's data they used |
| Residency | EU or US, matching your warehouse region; nothing in the Fabric is used to train models |
| Plan availability | Enterprise |
How it fits
How Data Fabric fits the platform
One identity model, one policy layer, one log. The other components and the trust model that binds them.
Questions
Questions about Data Fabric
Zero-copy sharing lets a consumer query data in a provider's warehouse without extracting or duplicating it. Meridian's agents read your Snowflake, Databricks, or BigQuery data this way, so the warehouse's security policies apply and there is no second copy to govern.
In your warehouse, where it already lives, for zero-copy sources. For connector-fed sources, in an Apache Iceberg lakehouse in the region you choose, in storage you can inspect. Meridian does not maintain a separate copy of your data for its own purposes and does not train on it.
More than 3,000 applications, including the common HRIS, ATS, payroll, ERP, CLM, banking, expense, ticketing, and collaboration platforms. Connectors are generic categories on this site; your account team can confirm coverage for your specific stack.
Yes. Iceberg is an open table format. Your BI tools, notebooks, and warehouses can read the same tables the agents read, with the same permissions.
Get started
Put the first agent to work this quarter.
Start with one workflow, one approver, and one number to move. Most design partners were live in five weeks.