live wire
▸Snowflake agent evals can pin judge versions; unpinned metrics may roll forward and shift scoresSnowflake▸ThoughtSpot makes Spotter 3 default for new instances; it requires download rights and always shares query data with an LLMThoughtSpot documentation▸MotherDuck adds image classification and object location to prompt_jev(), returning typed SQL outputs with confidence scoresMotherDuck▸BigQuery conversational analytics adds Montréal and Toronto; mixed-region agents still default to globalGoogle Cloud Documentation▸Fabric enables Azure OpenAI Copilot access by default, but OpenAI-subprocessor and cross-region processing/storage switches remain offMicrosoft Learn▸Snowflake Cortex Agents now continue by default when named tools are inaccessible, emitting warnings instead of failing the runSnowflake documentation▸Databricks Genie One can use group memberships to tailor answers, while Unity Catalog remains the access-control boundaryDatabricks documentation▸Fabric data agents can query Power BI models with Read alone; Build and workspace access are unnecessaryMicrosoft Learn▸Fabric data-agent Git configs expose column flags, but selecting a table still includes every columnMicrosoft Learn▸Google Data Agent Kit offers remote Spark kernels, but notebook exports can still move results into local Pandas or ArrowGoogle Cloud Documentation▸Cube says production analytics agents need explicit partial-result states, caller-filtered model search and durable artifactsCube▸Thinking Machines names Bridgewater behind ReViSQL-K2.6, citing 92.97% SC-16 accuracy at 15% of frontier-model costThinking Machines Lab▸AWS says a mismatched data/tool/prompt release cut its Field Sales agent pass rate from 78% to 38% in pre-productionAWS Artificial Intelligence Blog▸NL2SQL Multi-Agent ships evaluation artifacts across BIRD, CSpider, Archer and KDD DataSpace—but no software licenseGitHub▸Snowflake agent evals can pin judge versions; unpinned metrics may roll forward and shift scoresSnowflake▸ThoughtSpot makes Spotter 3 default for new instances; it requires download rights and always shares query data with an LLMThoughtSpot documentation▸MotherDuck adds image classification and object location to prompt_jev(), returning typed SQL outputs with confidence scoresMotherDuck▸BigQuery conversational analytics adds Montréal and Toronto; mixed-region agents still default to globalGoogle Cloud Documentation▸Fabric enables Azure OpenAI Copilot access by default, but OpenAI-subprocessor and cross-region processing/storage switches remain offMicrosoft Learn▸Snowflake Cortex Agents now continue by default when named tools are inaccessible, emitting warnings instead of failing the runSnowflake documentation▸Databricks Genie One can use group memberships to tailor answers, while Unity Catalog remains the access-control boundaryDatabricks documentation▸Fabric data agents can query Power BI models with Read alone; Build and workspace access are unnecessaryMicrosoft Learn▸Fabric data-agent Git configs expose column flags, but selecting a table still includes every columnMicrosoft Learn▸Google Data Agent Kit offers remote Spark kernels, but notebook exports can still move results into local Pandas or ArrowGoogle Cloud Documentation▸Cube says production analytics agents need explicit partial-result states, caller-filtered model search and durable artifactsCube▸Thinking Machines names Bridgewater behind ReViSQL-K2.6, citing 92.97% SC-16 accuracy at 15% of frontier-model costThinking Machines Lab▸AWS says a mismatched data/tool/prompt release cut its Field Sales agent pass rate from 78% to 38% in pre-productionAWS Artificial Intelligence Blog▸NL2SQL Multi-Agent ships evaluation artifacts across BIRD, CSpider, Archer and KDD DataSpace—but no software licenseGitHub
nl2sql.ai
newsMCP compatibility

GoodData’s MCP server stops advertising every tool up front

Tool search becomes the default on September 17, changing discovery behavior for clients that inspect tools/list before deciding what they can call.

MCP server discovery changes from full upfront listing to search-first access.
Side by side: what changed
By The News Desk· Sep 17, 2026the quick take — two AI hosts go live when you do

GoodData changed the default discovery behavior of its MCP server on September 17. Instead of returning every available capability in the initial MCP tool list, the server now exposes search_tools and call_tool as the primary discovery path. The underlying tools have not been removed and can still be invoked directly by name. GoodData release notes

The change matters most for integrations that treat tools/list as a complete capability inventory. GoodData explicitly warns that those integrations may need adjustment: clients should either search for the appropriate tool and then call it through call_tool, or invoke known tools directly by name. GoodData says off-the-shelf clients with tool-search support should continue working without changes. GoodData release notes

Why the tool list is changing

GoodData’s MCP documentation describes a split between always-on core tools and additional tools that clients discover on demand. Tool search matches a client’s intent against tool names and descriptions, allowing the server to offer a larger capability set without loading every definition into the initial exchange. The same discovery system applies to GoodData’s manually defined tools and generated wrappers for selected GoodData v1 API operations. GoodData MCP documentation

That design also creates a clear compatibility boundary. GoodData says clients without tool-search support see only the always-on core tools; dynamically generated tools are available only to clients that support discovery. In practice, a client may still establish an MCP connection while exposing only part of the server’s capability set. GoodData MCP documentation

The available surface is broader than conversational analytics alone. GoodData documents tools for workspace metadata, governed metric creation, data-source scanning and logical-model generation, alerts, scheduled exports, AI search, and execution of saved visualizations. Generated API tools are workspace-scoped and exclude PUT, PATCH, DELETE, action endpoints, and organization- or admin-scoped operations; they also enforce the permissions of the underlying GoodData API. GoodData MCP documentation

What operators should test

Teams running custom MCP clients should verify two separate behaviors: whether the client supports tool search, and whether any orchestration logic assumes that the first tools/list response is exhaustive. A successful connection is not enough to prove that all GoodData operations remain discoverable.

GoodData had already offered the new behavior through an Early Access setting called Extended MCP Tools. With the September 17 default now in effect, the practical migration check is straightforward: search for a non-core capability, call the returned tool, and separately test any direct-by-name calls the application depends on. GoodData release notes

sources

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.

    more from the desk