live wire
▸Qwen2.5-3B GRPO adapter posts 62.67% Spider dev execution accuracy, 1.8 points above base; base leads execution successHugging Face model card▸NextGen Healthcare opens N.I.A. analytics agent to early adopters; all Insights clients get it in early NovemberNextGen Healthcare▸Overseas Tax Agent posts a 200-case synthetic eval; tax math is deterministic, but SQL and RAG tools remain stubsGitHub repository▸Databricks says Genie rose from 32% to 90%+ on an internal benchmark; task count and artifacts are undisclosedDatabricks▸Tableau says agent traces and human corrections continually reshape its runtime knowledge graphTableau▸ThoughtSpot urges separate read/write permissions and explicit approval for production-impacting agent actionsThoughtSpot▸SchemaGate v1.11.1 drops sentence-like notes after complete SQL; 28,785-reply replay changed one trailing commentSchemaGate▸Teradata adds agent-aware ABAC, identity reduction and immutable backups for enterprise AI dataTeradata▸Huel says it moved staff to Spotter-first analytics after a direct Snowflake-to-Claude trial lacked explainabilityThoughtSpot customer essay▸KG mediation reaches 89.47%/90.29% F1 on 158 enterprise prompts; only 30 receive preliminary end-to-end trialsECIS 2026 paper▸MotherDuck says hiding query_rw is insufficient for MCP read-only mode; other write/delete tools need separate reviewMotherDuck▸Oracle AIDP adds progressively disclosed Agent Skills with optional declared Python entrypoints for high-code agentsOracle▸ThoughtSpot says Spotter 3 may retry forecasts with a different model after evaluating confidence intervalsThoughtSpot▸AtScale’s preliminary BIRD-Interact comparison uses best of 3 semantic-layer runs against a single control runAtScale▸Qwen2.5-3B GRPO adapter posts 62.67% Spider dev execution accuracy, 1.8 points above base; base leads execution successHugging Face model card▸NextGen Healthcare opens N.I.A. analytics agent to early adopters; all Insights clients get it in early NovemberNextGen Healthcare▸Overseas Tax Agent posts a 200-case synthetic eval; tax math is deterministic, but SQL and RAG tools remain stubsGitHub repository▸Databricks says Genie rose from 32% to 90%+ on an internal benchmark; task count and artifacts are undisclosedDatabricks▸Tableau says agent traces and human corrections continually reshape its runtime knowledge graphTableau▸ThoughtSpot urges separate read/write permissions and explicit approval for production-impacting agent actionsThoughtSpot▸SchemaGate v1.11.1 drops sentence-like notes after complete SQL; 28,785-reply replay changed one trailing commentSchemaGate▸Teradata adds agent-aware ABAC, identity reduction and immutable backups for enterprise AI dataTeradata▸Huel says it moved staff to Spotter-first analytics after a direct Snowflake-to-Claude trial lacked explainabilityThoughtSpot customer essay▸KG mediation reaches 89.47%/90.29% F1 on 158 enterprise prompts; only 30 receive preliminary end-to-end trialsECIS 2026 paper▸MotherDuck says hiding query_rw is insufficient for MCP read-only mode; other write/delete tools need separate reviewMotherDuck▸Oracle AIDP adds progressively disclosed Agent Skills with optional declared Python entrypoints for high-code agentsOracle▸ThoughtSpot says Spotter 3 may retry forecasts with a different model after evaluating confidence intervalsThoughtSpot▸AtScale’s preliminary BIRD-Interact comparison uses best of 3 semantic-layer runs against a single control runAtScale
nl2sql.ai
releaseNEW TOOL

Showboard gives coding agents a read-only SQL path to dashboards—but its benchmark is only a starting point

The new MIT-licensed MCP server turns PostgreSQL, SQLite and spreadsheet queries into inspectable boards, while its cost and speed claims remain narrowly author-tested.

By The Tools Desk· Oct 24, 2026the quick take — two AI hosts go live when you do

A dashboard tool, not another chat interface

Showboard is a new MIT-licensed MCP server that lets coding agents produce dashboards from PostgreSQL, SQLite, Excel and CSV data. Instead of asking the model to write a bespoke analysis script and HTML page, the agent sends SQL and a layout through MCP; Showboard runs the query and renders charts, tables and headline metrics in a browser, an application or a terminal pane. The repository names Claude Code, Codex, Cursor and Gemini CLI among its supported clients. (Showboard README)

That architecture makes the project notable for the NL2SQL directory. Showboard does not replace the model that writes SQL. It narrows what happens after generation: PostgreSQL queries run inside a read-only transaction with row and time limits, while SQLite files and spreadsheets are opened read-only. Every rendered element can expose its backing query, giving users a direct path from a number on screen to the SQL that produced it. (Showboard README)

The first public release line is already packaged beyond source code. Version 0.4.1, published October 22, includes a 2.9 MB MCPB extension for the Claude desktop app and adjusts the Claude Code plugin’s startup behavior for Anthropic’s plugin directory. (Showboard 0.4.1 release)

The benchmark is useful—and narrow

The project’s author reports a 24-answer comparison using Claude Code with Opus 5.5 on one spreadsheet dataset. Across those runs, Showboard cost $1.78 versus $4.77 for the agent working alone, used 36,659 versus 128,929 output tokens, and took 7.5 versus 24.3 minutes. The repository includes the questions and rerun instructions, which is better evidence than an unsupported efficiency claim. (Showboard benchmark)

But those numbers should not be generalized into a universal 2.7× saving. The project explicitly says the comparison covers two runs on one dataset, and other agents have not yet been measured end to end. It also notes that the standalone Claude pages were more individually designed than Showboard’s standardized boards. (Showboard README)

The practical reason to test Showboard is therefore not the headline multiplier. It is the operating boundary: query execution is constrained, the visualization path is reusable, and the SQL remains inspectable. Teams evaluating it should rerun the supplied benchmark on their own schemas, models and recurring questions, then separately check query correctness and permission behavior. The current evidence shows a credible new tool with a reproducible test harness—not an independently validated performance result.

sources

Filed by The Tools 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