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
newsEnterprise release

SAP’s BTP administration agent inherits the operator’s identity—not a superuser token

A new remote MCP server lets compatible assistants inspect, provision and manage SAP BTP resources while remaining bound to the signed-in user’s existing permissions.

Assistant-to-BTP access diagram with identity-limited admin paths.
AI-generated diagram
By The News Desk· Aug 31, 2026the quick take — two AI hosts, this story only

SAP has made its MCP server for SAP Business Technology Platform administration publicly available as a remote service. The release gives compatible AI clients a natural-language route into BTP account administration without asking operators to host a proxy application or compile server code. SAP positions it as a way to reduce the chain of cockpit navigation and CLI calls needed for routine discovery, auditing and provisioning tasks. SAP’s release post

The important boundary is identity

The useful design choice is not that an assistant can call infrastructure APIs. It is that SAP says authentication uses the authorization-code flow through its default identity provider and propagates the signed-in user’s principal. The MCP server therefore operates with that person’s existing IAM policies and role collections rather than receiving a broader standing credential. In SAP’s examples, the same interface can check entitlements, audit a user’s roles across production subaccounts, or create a QA subaccount patterned after staging—but only within the operator’s own authority. SAP’s release post

That boundary matters because an administrative agent may dynamically sequence several API calls from one instruction. The language model can choose the tools required to satisfy an intent, while BTP’s authorization system still decides which actions are permitted. It is a more defensible pattern than handing an agent an all-powerful service credential and relying on prompt instructions to limit its behavior. SAP’s release post

One endpoint, four domains

SAP’s documentation says the server aggregates four administration domains—accounts, security, services and connectivity—behind one endpoint. Compatible clients discover the available operations and select tools based on the request; administrators do not configure each operation individually. SAP names Claude Code, GitHub Copilot in VS Code and OpenCode as examples of compatible clients, while saying any MCP-standard client can connect. SAP Help Portal

The service also handles regional routing after a single connection, according to SAP. That gives an operator one conversational control surface across BTP regions without juggling regional endpoints. The public release post supplies a remote MCP URL and client ID, while directing administrators to the Help Portal for current connection parameters and supported clients. SAP’s release post

What teams should test

Before treating the interface as an operational shortcut, teams should test the permission boundary with deliberately over-scoped requests: ask a low-privilege user’s agent to inspect an unauthorized subaccount, assign a forbidden role and provision an unapproved entitlement. The key acceptance criterion is not whether the model refuses politely; it is whether the platform denies the underlying tool calls. SAP’s principal-propagation design gives teams a concrete control to verify, while the public endpoint removes the deployment work that often obscures that test. SAP’s release post

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.