diff --git a/airbyte/_direct_connectors/actions.py b/airbyte/_direct_connectors/actions.py
index e9b4dd8fd..eb3fa3282 100644
--- a/airbyte/_direct_connectors/actions.py
+++ b/airbyte/_direct_connectors/actions.py
@@ -26,8 +26,9 @@ class AgentReadAction(str, Enum):
     """Connector actions that only read data.
 
     The `search` action is the connector's native API search, parallel to `get` and `list`.
-    The `sql_select` action runs one read-only SQL statement (or `SHOW TABLES`) on the query
-    engine behind a destination connector. Pass `sql` and `sql_dialect` (and optionally
+    The `sql_select` action runs one read-only SQL statement on the query engine behind a
+    destination connector: a `SELECT`/`WITH` query or the `SHOW TABLES`, `SHOW DATABASES`,
+    or `DESCRIBE TABLE <table>` macro. Pass `sql` and `sql_dialect` (and optionally
     `dry_run`) in `api_args`; `entity_type` is ignored for this action.
 
     The `download` action is deliberately absent even though it reads: it returns a binary
diff --git a/airbyte/cloud/connectors.py b/airbyte/cloud/connectors.py
index e1ae7693f..5dbda6bf0 100644
--- a/airbyte/cloud/connectors.py
+++ b/airbyte/cloud/connectors.py
@@ -461,7 +461,10 @@ def execute_sql_query(
         cursor: str | None = None,
         dry_run: bool = False,
     ) -> ExternalApiExecuteResult:
-        """Run a read-only SQL `SELECT` (or `SHOW TABLES`) through SQL passthrough.
+        """Run a read-only SQL statement through SQL passthrough.
+
+        Accepted statements are a single `SELECT`/`WITH` query or the `SHOW TABLES`,
+        `SHOW DATABASES`, or `DESCRIBE TABLE <table>` macro.
 
         `sql_dialect` defaults to the dialect registered for this connector's definition.
         `dry_run=True` validates the statement and returns the result columns without
diff --git a/airbyte/mcp/cloud.py b/airbyte/mcp/cloud.py
index 05124c5f2..7dee6b512 100644
--- a/airbyte/mcp/cloud.py
+++ b/airbyte/mcp/cloud.py
@@ -1818,7 +1818,12 @@ def execute_external_sql_query(
     ],
     sql: Annotated[
         str,
-        Field(description="The read-only SQL statement to run, for example `SHOW TABLES`."),
+        Field(
+            description=(
+                "The read-only SQL statement to run: a `SELECT`/`WITH` query or the "
+                "`SHOW TABLES`, `SHOW DATABASES`, or `DESCRIBE TABLE <table>` macro."
+            )
+        ),
     ],
     sql_dialect: Annotated[
         str | None,
@@ -1869,12 +1874,14 @@ def execute_external_sql_query(
 
     Only SQL-passthrough destinations (Snowflake/BigQuery) support this tool.
 
-    Run `SHOW TABLES` first to discover tables; `sql_dialect` defaults to the
-    destination's registered dialect. Use `dry_run=True` with
-    `SELECT * FROM <table> LIMIT 1` to discover a table's columns.
+    Run `SHOW TABLES` first to discover tables and `SHOW DATABASES` to see the
+    databases the destination is scoped to; `sql_dialect` defaults to the
+    destination's registered dialect. Use `DESCRIBE TABLE <table>` to return a
+    table's columns without reading rows.
 
-    `SHOW TABLES` is the only non-`SELECT` statement accepted and takes no
-    `LIMIT`; add a `LIMIT` to every `SELECT`. On Snowflake prefer unquoted
+    Accepted statements are a single `SELECT`/`WITH` query plus the `SHOW TABLES`,
+    `SHOW DATABASES`, and `DESCRIBE TABLE <table>` macros. The `SHOW` macros take
+    no `LIMIT`; add a `LIMIT` to every `SELECT`. On Snowflake prefer unquoted
     identifiers: double-quoting makes them case-sensitive.
     """
     connector = _get_cloud_workspace(ctx, workspace_id).get_connector(connector_id)