live wire
IBM makes watsonx Orchestrate AgentOps, custom LLM judging and Bedrock-agent discovery generally availableIBMSchemaGate 0.1.45 fixes broken Oracle ADB wallet connections and an OCI stack pinned 28 releases behindSchemaGatePDI’s Amazon Quick procurement agent grounds spend answers in vendor, category and contract contextAWS Business Intelligence BlogBigQuery’s ML.METRICS example returns 0.84 accuracy but 0.30 macro-F1 on the same 100-row classification queryGoogle Cloud BigQuery docsSchemaGate 0.1.44 auto-selects sentence embeddings, lifting bundled-schema retrieval from 90/98 to 93/98SchemaGateSchemaGate 0.1.43 adds read-only SQL execution with per-principal table checks—and documents unauthenticated client assertionsSchemaGateDatabox adds reusable AI Analyst Skills with personal/company scope, auto-matching and marketplace installsDataboxFabric previews an AI builder for data-agent instructions, source guidance and example queriesMicrosoft FabricDatabricks trains data-agent retriever to stop early or spend bounded extra search steps, reporting 5.8-second latencyDatabricksThoughtSpot adds SpotterCode coding agent to its Visual Embed PlaygroundThoughtSpotLongMemEval-S audit: 67–73% of restore-fixable 80k-budget errors came from evicted evidence under three policiesarXivSchemaGate 0.1.42 adds dimension-aware retrieval and fixes complex multi-table SQL promptsSchemaGateSnowflake agent toolsets can silently drop inherited tools when callers lack accessSnowflake DocumentationLooker’s VS Code extension reaches GA with MCP-assisted LookML generation, editing and validationGoogle Cloud Looker release docsIBM makes watsonx Orchestrate AgentOps, custom LLM judging and Bedrock-agent discovery generally availableIBMSchemaGate 0.1.45 fixes broken Oracle ADB wallet connections and an OCI stack pinned 28 releases behindSchemaGatePDI’s Amazon Quick procurement agent grounds spend answers in vendor, category and contract contextAWS Business Intelligence BlogBigQuery’s ML.METRICS example returns 0.84 accuracy but 0.30 macro-F1 on the same 100-row classification queryGoogle Cloud BigQuery docsSchemaGate 0.1.44 auto-selects sentence embeddings, lifting bundled-schema retrieval from 90/98 to 93/98SchemaGateSchemaGate 0.1.43 adds read-only SQL execution with per-principal table checks—and documents unauthenticated client assertionsSchemaGateDatabox adds reusable AI Analyst Skills with personal/company scope, auto-matching and marketplace installsDataboxFabric previews an AI builder for data-agent instructions, source guidance and example queriesMicrosoft FabricDatabricks trains data-agent retriever to stop early or spend bounded extra search steps, reporting 5.8-second latencyDatabricksThoughtSpot adds SpotterCode coding agent to its Visual Embed PlaygroundThoughtSpotLongMemEval-S audit: 67–73% of restore-fixable 80k-budget errors came from evicted evidence under three policiesarXivSchemaGate 0.1.42 adds dimension-aware retrieval and fixes complex multi-table SQL promptsSchemaGateSnowflake agent toolsets can silently drop inherited tools when callers lack accessSnowflake DocumentationLooker’s VS Code extension reaches GA with MCP-assisted LookML generation, editing and validationGoogle Cloud Looker release docs
nl2sql.ai
guidePRACTITIONER GUIDE

Oracle’s OAC MCP demo makes permission changes part of the analytics-agent review

A new workbook-automation walkthrough shows why teams should separate content generation from authorization changes and inspect both before release.

Filtered workbook copies versus separate permission approval in Oracle Analytics Cloud.
AI-generated illustration
By The News Desk· Sep 10, 2026the quick take — two AI hosts go live when you do

Oracle’s newest Oracle Analytics Cloud walkthrough moves the analytics agent beyond answering questions or drafting SQL. The demonstrated workflow starts with one workbook, creates filtered copies for AMERICAS, APAC and EMEA, places them in matching shared folders, and then assigns matching custom application roles—all from natural-language prompts issued through Codex and OAC’s MCP server. Oracle’s walkthrough says other external AI clients are supported as well.

That makes this a permissions workflow, not merely a content-generation demo. Oracle explicitly recommends reviewing and validating automated work, particularly changes to catalog structure and permissions. Its MCP documentation adds the sharper security boundary: MCP calls use an OAuth 2.0 access token, execute with the permissions of the user who obtained it, and can expose unintended data if that user can reach content that should not be shared with an LLM.

The safe rollout is a two-stage job

The operational pattern should therefore be split into two approvals. First, let the agent create or duplicate the filtered workbooks and proposed folder structure. A reviewer should confirm that each copy has the intended filter and destination. Only then should a separately approved step assign application roles. This separation is our deployment recommendation, derived from Oracle’s own instruction to validate catalog and permission changes and from the documentation’s same-as-user authorization model. Oracle’s example performs workbook creation first and role assignment in a subsequent prompt.

Before enabling that second step, test the assumption hidden in the demo: folder names and role names match exactly. The example succeeds because AMERICAS, APAC and EMEA are both folder names and custom role names. The walkthrough does not describe what happens when a role is missing, duplicated or ambiguously named, so production teams should make those failure cases part of acceptance testing rather than infer safe behavior.

A five-point review checklist

Based on the published contract, the minimum review should be:

  1. Constrain the token holder. Give the MCP session only the OAC access required for the target catalog operation; Oracle says every call inherits that user’s permissions. Documentation
  2. Preview content separately from authorization. Confirm workbook filters and destinations before any role change. Walkthrough
  3. Compare permissions before and after. Treat the agent’s result as a proposed catalog change, not proof that access is correct; Oracle tells administrators to review and validate the automation. Walkthrough
  4. Exercise naming failures. Test absent, duplicate and near-match roles because the published success case depends on exact name correspondence. Walkthrough
  5. Instrument the workflow. Oracle recommends evaluations and observability for agents using the MCP server as the protocol and tools evolve. Documentation

The useful shift is not that a chat client can copy three reports. It is that enterprise analytics agents can now change the catalog and its audience in one workflow. Oracle’s own guidance points to the right control: keep the automation, but make permission assignment an inspectable, separately approved deployment step.

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.