Governed, fresh, and traceable: tool by tool.
Agent Views are exposed through the Dagen MCP server so Claude, Copilot, or any MCP-compatible client can discover, describe, and query them safely. Each property maps to tools documented on the MCP reference.
Governed
Every tool call is scoped to the Agent Views the authenticated user can already see: the server enforces this before a query runs, not after. Row- and column-level permissions apply before an agent ever sees a record. What a user can’t see, their agent can’t either.
describe_agent_view query_agent_view
Fresh
get_pipeline_status reports the health of the pipelines feeding an Agent View: freshness, last run status, and any open remediations. list_agent_views returns the last-refreshed time for each view, so pipeline freshness is something an agent can check, not assume.
get_pipeline_status list_agent_views
Traceable
get_lineage traces a field or table back through its transformations to the original source. The receipt: source table, agent view, and governance tag, comes back with every answer.
get_lineage source agent_view governed
list_agent_views
Lists the Agent Views the current user is authorized to reach.
describe_agent_view
Returns the definitions, columns, and access policy for one Agent View.
query_agent_view
Runs a governed query against an Agent View and returns rows.
get_lineage
Traces a field or table back through its transformations to the original source.
get_pipeline_status
Reports the health of the pipelines feeding an Agent View.
A simplified response for query_agent_view, from the MCP reference. The receipt comes back with every answer.
{ "answer": "...",
"source": "fct_opportunity",
"agent_view": "revenue_360",
"governed": true }
Keep Snowflake or Databricks. Keep their governance.
Dagen does not ask you to rip and replace the catalog, the warehouse, or the orchestrator. It sits above what you already run and makes that estate usable by agents.
Snowflake
Snowflake stays exactly where it is: the platform your team trusts. Dagen connects to Snowflake with credentials you scope and govern. Pipelines deploy in your environment and respect your existing roles and grants. Agent Views on your warehouse data are served over MCP, so Claude and any MCP-compatible agent query governed Snowflake data.
Databricks
Your lakehouse stays on Databricks. Pipelines respect your Unity Catalog governance: your permissions, your lineage, your audit requirements. The agentic runtime monitors and heals lakehouse flows while pipelines inherit the Unity Catalog governance you already enforce. Agent Views sit on your Delta and Iceberg tables and are served over MCP, so AI agents consume governed lakehouse data directly.
Agents query the view, not the raw warehouse.
A raw MCP gateway to tables still leaves cryptic columns, missing definitions, and no attachment of policy and lineage to the answer.
A catalog entry describes data for a human to read. An Agent View is queryable: an agent calls it directly over MCP and gets rows back, governed the same way a person’s access would be.
An Agent View is built on top of a database view, but a database view has no definitions, access policy, or lineage attached. An Agent View carries all three, so an agent can use it safely, no human translation required.
Same pattern for Iceberg on S3 and hybrid estates.
Reuse over rip-and-replace. Open table formats and AI-in-front-of-data are why pipelines need a control plane; they are not a mandate to abandon Snowflake or Databricks. Many teams adopt Iceberg on S3 (AWS-first) while keeping the warehouse for the workloads that belong there.
Jobs still land on the engines you already own. Glue, EMR, Athena, Spark, DuckDB, Snowflake, Databricks, interchangeable behind the map. Customers keep every runtime of their own.
Governed agent access to Snowflake and Databricks, answered.
How do I give AI agents governed, fresh, traceable access to our Snowflake or Databricks estate?+
Keep Snowflake or Databricks. Dagen maps the tables, models, and pipelines you already have, applies the access policy you already trust, and publishes Agent Views over MCP. Agents query the view, not the raw warehouse, so governed enterprise data reaches agents over MCP with definitions, lineage, and freshness attached. Same pattern for Iceberg on S3 and hybrid estates. You do not have to migrate off the warehouse; many teams adopt Iceberg on S3 (AWS-first) while Snowflake stays for the workloads that belong there.
Does Dagen replace Unity Catalog or Snowflake governance?+
No. Dagen does not ask you to rip and replace the catalog, the warehouse, or the orchestrator. On Databricks, pipelines run under your Unity Catalog governance, and Dagen records lineage alongside the permissions you already enforce. On Snowflake, you scope the credentials Dagen uses, and pipelines write to your databases under the roles and grants you already enforce.
How do agents know the data is fresh?+
Agents call get_pipeline_status, which reports the health of the pipelines feeding an Agent View: freshness, last run status, and any open remediations. list_agent_views also returns the last-refreshed time for each view.
Can an agent see data its user can’t?+
No. Every tool call is scoped to the Agent Views the authenticated user can already see: the server enforces this before a query runs, not after. What a user can’t see, their agent can’t either.
How is an Agent View different from a data catalog entry?+
A catalog entry describes data for a human to read. An Agent View is queryable: an agent calls it directly over MCP and gets rows back, governed the same way a person’s access would be.
Is there an MCP server for a Snowflake or Databricks estate?+
Yes. Keep Snowflake or Databricks. Dagen maps the tables, models, and pipelines you already have, applies the access policy you already trust, and serves Agent Views over MCP. Agents query the view, not the raw warehouse, so a Snowflake or Databricks estate becomes agent-ready without rip-and-replace. Same pattern for Iceberg on S3 and hybrid estates.
Explore the control plane
Data control plane
Map the estate. Govern access. Serve Agent Views over MCP.
ProductAgent Views
The governed interface between your data and every AI agent.
ProductMCP
Governed enterprise data for AI agents over MCP, the hub.
ReferenceMCP reference
Tools and resources the Dagen MCP server exposes.
Ready to take control of your enterprise data?
Map the estate. Govern access. Serve Agent Views over MCP, from the data you already run.