live wire
▸Snowflake’s Cortex Agent quickstart routes metrics and transcripts with prompt rules—but documents no authorization boundarySnowflake▸Oracle Select AI generates SQL/PGQ for property graphs—but errors if one AI profile mixes graphs with tables or viewsOracle▸Domo’s data team standardizes MCP builds with shared scaffolding; beta toolkits inherit user permissionsDomo▸Databricks proposes five outcome signals for agentic BI—and says dashboard counts and license usage miss decision qualityDatabricks▸One vibe-coded database app needed 56+ human test hours and three repair rounds despite AI Studio’s verification claimsMancas & Mancas / arXiv▸BigQuery AI.IF’s default cost mode trains a local model for most rows—and drops row-error thresholdsGoogle Cloud▸AWS’s multi-tenant data-agent pattern gives the broker Lambda no data permissions; Cognito roles carry Lake Formation scopeAWS Big Data Blog▸Amazon OpenSearch’s optimized engine adds nested agent traces, but AWS reserves it for log analytics—not vector searchAWS▸Two interpretation-based provenance semantics cannot express SQL bag difference, new theorem showsKolaitis / arXiv▸Databricks Feature Views reuse one definition for training and live scoring, but online lookup still returns latest available stateDatabricks▸Databricks Genie’s Slack and Teams previews default channel answers to public, even when viewers lack source-data accessDatabricks▸Snowflake’s two-tool GraphRAG agent beat a seven-tool graph setup 70% to 60% on 22 ontology questionsSnowflake Engineering▸Semantix opens an Apache-2.0 semantic-first data-agent stack, but labels the project WIP and has no release or adoption yetGitHub / icosmos-space▸Confluent defines Streamhouse as shared current state for agents and apps, but says broader semantics and governance are still to comeConfluent▸Snowflake’s Cortex Agent quickstart routes metrics and transcripts with prompt rules—but documents no authorization boundarySnowflake▸Oracle Select AI generates SQL/PGQ for property graphs—but errors if one AI profile mixes graphs with tables or viewsOracle▸Domo’s data team standardizes MCP builds with shared scaffolding; beta toolkits inherit user permissionsDomo▸Databricks proposes five outcome signals for agentic BI—and says dashboard counts and license usage miss decision qualityDatabricks▸One vibe-coded database app needed 56+ human test hours and three repair rounds despite AI Studio’s verification claimsMancas & Mancas / arXiv▸BigQuery AI.IF’s default cost mode trains a local model for most rows—and drops row-error thresholdsGoogle Cloud▸AWS’s multi-tenant data-agent pattern gives the broker Lambda no data permissions; Cognito roles carry Lake Formation scopeAWS Big Data Blog▸Amazon OpenSearch’s optimized engine adds nested agent traces, but AWS reserves it for log analytics—not vector searchAWS▸Two interpretation-based provenance semantics cannot express SQL bag difference, new theorem showsKolaitis / arXiv▸Databricks Feature Views reuse one definition for training and live scoring, but online lookup still returns latest available stateDatabricks▸Databricks Genie’s Slack and Teams previews default channel answers to public, even when viewers lack source-data accessDatabricks▸Snowflake’s two-tool GraphRAG agent beat a seven-tool graph setup 70% to 60% on 22 ontology questionsSnowflake Engineering▸Semantix opens an Apache-2.0 semantic-first data-agent stack, but labels the project WIP and has no release or adoption yetGitHub / icosmos-space▸Confluent defines Streamhouse as shared current state for agents and apps, but says broader semantics and governance are still to comeConfluent
nl2sql.ai
releaseRelease

SchemaGate 0.1.43 can execute SQL—but its caller identity is still an assertion

The MCP server now closes the loop from schema selection to read-only query execution, while explicitly warning that principal and role claims are not authenticated.

SchemaGate now runs SQL, but principal and role claims are still only asserted.
Side by side: what changed
By The News Desk· Sep 13, 2026the quick take — two AI hosts go live when you do

SchemaGate 0.1.43 moves the project from helping an agent choose database context to executing the resulting query. The release adds a run_query MCP tool that accepts one read-only SELECT and returns rows, closing the gap after select_schema supplied DDL and the client’s model generated SQL. The project also keeps an answer() path for agents with their own configured model; without one, it returns the schema selection and points the caller to run_query instead of failing. Source: SchemaGate v0.1.43 release

Execution gets three gates

Before a query reaches the database, SchemaGate says it checks that the request contains one statement, that the statement is a read, and that every referenced table is visible to the current principal. The release says the table test covers names inside joins and common table expressions, rather than only the most obvious table token. It also says an earlier authorized call does not carry authorization into a later query. Source: SchemaGate v0.1.43 release

The refusal behavior is deliberately conservative. Anonymous callers, callers with the wrong role and queries that reference unknown tables are rejected. SchemaGate does not tell the caller whether an unknown name is restricted, misspelled or absent from the reflected schema. That avoids turning the error path into a catalog-discovery side channel, although the release does not claim this is a complete database-security boundary. Source: SchemaGate v0.1.43 release

Database errors are returned in the database’s own wording so a model can attempt a repair. That is useful for an agent loop, but operators should still treat those messages as potentially sensitive because they can expose details about the database implementation or query shape. The release documents the pass-through behavior; it does not say errors are redacted. Source: SchemaGate v0.1.43 release

The trust boundary is the deployment decision

The most important caveat is explicit: principal and roles are asserted by the MCP client, and the server believes them. There is no token or session check, and a shared secret is not implemented in this release. The project describes that arrangement as appropriate over stdio or behind an authenticating layer, and inappropriate on an open port. Source: SchemaGate v0.1.43 release

That makes v0.1.43 a meaningful execution release, not a standalone authorization system. Teams adopting it should place authentication in front of any network exposure and bind verified user identity to the principal and role fields before relying on the table checks. The release reports 858 tests plus a property suite, but publishes no production deployment evidence or workload benchmark, so those controls still need testing against each organization’s schemas, drivers and identity path. Source: SchemaGate v0.1.43 release

Filed by The News Desk. Corrections: desk@nl2sql.ai · Our standards →

comments · 0

    Comments are moderated before they appear. Your email is used once to confirm it is you — never shown, never sold. Corrections and questions get an answer from the desk when we have one.