Live text-to-SQL apps

The interactive tools - connect a remote database or drop a file, then analyze and transform it in natural language. All open source.

• DATS-4 Database AI Suite - remote databasesThe flagship workbench: connect to any remote DB, upload CSV/TXT into tables, analyze and transform with natural language, Python charts and stats. Multi-model.
• BRIQ - In-Browser DuckDBDuckDB runs locally in your browser - process large CSVs, natural-language to SQL, charts. Data never leaves your machine.
• DuckIt - CSV to DuckDBConvert CSV/TSV (and Parquet) into a DuckDB database and get a shareable link.
• CSV ProcessorClean, merge and reshape CSV files with AI before they hit a database.
• Rapid Deploy - Deepseek R1 (Flowise UI)The text-to-SQL agent on a Flowise UI, powered by Deepseek R1. Connect to any database.
• Rapid Deploy - Quants Agent (Flowise UI)Quants, technicals and financials over a database, Flowise UI, PDF and web reports.

Database MCP - talk to your database from Excel or Claude

The read-only Database MCP lets an AI client (Claude in Excel, Claude Desktop, any MCP client) query your Postgres or DuckDB in plain English and pull the rows straight back. No key, no sign-up.

• Database Query MCP - referenceThe DuckDB and Postgres read-only SQL MCP server: what it does, how to connect, the safety model. No key.
• MCP server showcaseThe one-page card for the database MCP server, for humans and AI agents.
• Talk to Your Database from Excel - Postgres, DuckDB, via Claude + MCPThe walk-through: wire Claude in Excel to a live database over MCP.
• Talk to Your Database from Excel - Part 2The follow-up, deeper patterns.
• Claude in Excel with MCP Connector - talk to your backendsReach any backend from inside the sheet.
• MCP Guide for Claude in Excel - and any MCP clientThe general setup guide, start here to wire up MCP.

Custom GPTs connected to databases

Ready-made ChatGPT custom GPTs wired to live databases via custom actions. Source and OpenAPI schemas are embedded in each app page.

• REXDB GPT - ChatGPT connected to your databasesQuery Postgres / MySQL from ChatGPT with a hardened read-only backend.
• Cricket (ODI/T20) and Tour de France GPTA Custom GPT connected to 3 live sports databases - a worked example of the pattern.
• n8n Automation GPTChatGPT to n8n for database queries, automation and Google Apps Script.
• ChatGPT connected to your databases - deployment guideOne-click deployment instructions for AI coders.

Database infrastructure & production backends

How the data layer is provisioned and served at scale - the pieces behind the apps.

• Neon Database Proxy - instant Postgres provisioningProgrammatic Neon Postgres database/role creation and automated cleanup. Open source, hardened.
• Instant Database Setup for AI Apps, with NeonWhy and how the provisioning proxy works.
• A dashboard on 230M+ rows with DuckDB + FastAPIArchitecture for hundreds of millions of records at sub-second query.
• Full backend infrastructure for a DuckDB production systemThe end-to-end production setup.

Field Guide - deploying AI on databases (PDF)

A 50-page field guide from 15+ months of running the DATS-4 suite for live client operations. Practice, not theory: security hard rules, datamart construction, agent orchestration, and LLM cost management.

• Field Guide - Main (Aug 2025)The full 50-page guide.
• Costing - Qwen3 Max (Sep 2025)Cost/quality update.
• Costing - Claude Sonnet 4.5 (Sep 2025)Cost/quality update.
• Costing - GPT 5.1 / Gemini 3 Pro (Dec 2025)Latest cost/quality matrix.
• LLM Costing for Database AI Apps - live experienceThe write-up behind the costing PDFs.

Guides & deep dives (60+ posts)

A curated slice of the practitioner library - lessons, patterns and demos from building text-to-SQL agents on real, large data. The full set is linked at the end.

• Database AI, built for day-to-day work: five categories, ten micro-appsThe overview of the whole suite.
• Mistakes I Made Building Text-to-SQL Agents - 2025 learningsThe hard-won lessons.
• Try Text-to-SQL on Real Data - multi-million rows, GB+ filesA live demo on data that is actually big.
• Large File Upload for Database AI Text-to-SQL AppsA practical guide to the upload problem.
• DuckDB is not just fast SQL - it is Python, SQL and compression in one boxWhy DuckDB anchors so much of this.
• My Public MCP Server Got Hammered - security lessons from a bot attackWhat breaks when you expose an MCP to the internet.
• In analytics the SQL was always the easy part. Anthropic put a figure on itSep 2026. Their own evals: under 21% accuracy without skills, above 95% with them, same model and same warehouse. A practitioner note at the end.
• Have your AI analyze the cricket data: one pasted instructionAug 2026. The agent finds the service from the catalog, reads the spec, discovers the tables and runs the SQL.
• Both formats on both engines: Postgres and DuckDBAug 2026. A query no longer has to follow the format to a particular engine.
• All Database-AI & Text-to-SQL guidesThe full tag index, for humans and AI agents.