A self-hosted data analysis platform — structured recipe builder with AI-guided data preparation, and multi-database analytics. Open source. Runs on your infrastructure. Connects to PostgreSQL, MySQL, SQL Server, and MS Access.
Learn MoreStatistician's Workbench is a self-hosted web application that lets you connect to your databases, import files, build data preparation pipelines with a structured recipe editor or an AI assistant, and run analytics — all through a browser. Your data stays on your server. No cloud, no third-party processing, no data leaving your network.
Build data preparation pipelines step by step — chain stages like filter, join, aggregate, pivot, and impute. Each stage previews its output so you can see exactly what your data looks like at every step.
An AI assistant guides you through building a recipe — it sees your actual schema and data, then suggests which stages to use and how to configure them. Every suggestion is recorded alongside your decisions for a full audit trail. Connect to cloud AI providers or run a local LLM for total privacy.
Connect to PostgreSQL, MySQL, SQL Server, and MS Access. Browse schemas, preview tables, and run queries across all your databases from a single interface.
Import CSV, JSON, and Excel files with automatic type detection. Preview data before import. Map columns, set types, and load directly into your database.
Visualise results with bar charts, line charts, scatter plots, and histograms. Charts update as you refine your recipe — instant feedback on every change.
Open any dataset as a Jupyter notebook for advanced analysis. Python, R, and SQL kernels — write code, generate reports, and share results, all within the Workbench.
Full interface translation across multiple languages including Thai. Switch languages from the globe icon — no restart required.
Multi-user support with role-based access control. Group admins manage their own users. Fine-grained permissions on datasets, recipes, and database connections.
Runs on your own server — a standard Linux machine with Python and PostgreSQL. Deploy directly. No cloud dependency. Your infrastructure, your control.
The Workbench is a web application built with a Python backend (FastAPI) and a Next.js frontend. Deploy it on your server and access it through any modern browser:
| Step | What Happens |
|---|---|
| 1. Deploy | Clone the repository. Point it at a PostgreSQL database. The Workbench creates its own schema for user accounts, recipes, and datasets. |
| 2. Connect | Add your database connections — PostgreSQL, MySQL, SQL Server, or MS Access. The Workbench uses native drivers (psycopg2, pymysql, pymssql, pymsaccess) to connect directly to each. |
| 3. Import | Upload CSV, JSON, or Excel files. Preview the data, adjust column types, and load into your database — or keep them as virtual datasets that query the file directly via DuckDB. |
| 4. Build | Create a recipe using the structured editor — chain stages like Filter, Join, Aggregate, Pivot, and Impute — or use the AI assistant, which suggests stages and configuration based on your actual data. Each stage shows a live preview of its output. Every AI suggestion is recorded, giving you a full audit trail. |
| 5. Analyse | Run the recipe to produce a dataset. Chart it, export it to CSV/Excel, or open it as a Jupyter notebook for deeper analysis with Python or R. |
| Database | Driver | Features |
|---|---|---|
| PostgreSQL | psycopg2 | Full support — schema browsing, table preview, SQL querying, recipe execution, data import/export |
| MySQL / MariaDB | pymysql | Full support — schema browsing, table preview, SQL querying, recipe execution |
| SQL Server | pymssql | Full support — schema browsing, table preview, SQL querying, recipe execution. Feature-gated behind mssql_db licence flag |
| MS Access | pymsaccess | Schema browsing, table preview, SQL querying. Read-only access to .accdb and .mdb files |
| DB2 | Planned | IBM Db2 support — planned for the Business Edition. Schema browsing, table preview, SQL querying, recipe execution. |
| IMS / Mainframe | Planned | IBM IMS and other mainframe data sources — planned for the Business Edition. Enterprise data integration for legacy systems. |
The platform targets full support for 20 languages. The Workbench already supports 9 of them in the application UI, and this website will soon be updated with pages in all 20.
| Code | Language | Native Name |
|---|---|---|
en | English | English |
zh | Chinese | 中文 |
th | Thai | ไทย |
de | German | Deutsch |
ar | Arabic | العربية |
fa | Farsi | فارسی |
bn | Bengali | বাংলা |
pa | Punjabi | ਪੰਜਾਬੀ |
es | Spanish | Español |
ja | Japanese | 日本語 |
fr | French | Français |
pt | Portuguese | Português |
it | Italian | Italiano |
ko | Korean | 한국어 |
ru | Russian | Русский |
hi | Hindi | हिन्दी |
nl | Dutch | Nederlands |
tr | Turkish | Türkçe |
id | Indonesian | Bahasa Indonesia |
ur | Urdu | اردو |
Version 0.12.0 — in development. 0.11.0 (released 2026-07-24) delivered AI agents, chat, notebook import, mobile support, audio I/O, and MSSQL connectivity. 0.12.0 is the Recipe Redesign & Platform Polish milestone — a complete rebuild of the recipe system with pipeline editor, SQL execution pushdown across five database backends, and platform-wide translation compliance. 15 of 21 tasks complete.
| Milestone | Version | Theme | Status |
|---|---|---|---|
| 0.0.1 — Foundation | 0.9.0 | Dataset management, password policy, admin features | ✓ Released |
| 0.0.2 — Statistical Foundation | 0.10.0 | Core statistics, DuckDB engine, recipes, sharing, multilingual UI, database connections | ✓ Released |
| 0.0.3 — AI Agents & Collaboration | 0.11.0 | Chat agents, AI providers, notebook import, mobile support, audio I/O, MSSQL, Golden Screwdriver | ✓ Released |
| 0.0.4 — Recipe Redesign & Platform Polish | 0.12.0 | Pipeline editor, dataset transforms, SQL execution engine (5 backends), recipe sharing, translation compliance | 🔄 In Progress (15/21) |
| 0.0.5 — Production Readiness | 0.13.0 | Security audit, performance optimisation, activity logs, comprehensive testing, documentation, backup & recovery | 📋 Planned |
The Workbench is available in four editions. The base edition is open source. The other three are binary builds targeting less technical users who want an easy install:
| Product | License | Notes |
|---|---|---|
| Workbench | Open Source | The community edition. Full platform — PostgreSQL backend, recipe engine, statistics, charts, multilingual UI. Free to use, modify, and self-host. This is the core of the project. |
| Workbench-BE | Commercial | The Business Edition. Adds MSSQL and MS Access connectivity, Jupyter notebooks, and priority support. Annual licence per deployment. All editions include multi-user support with role-based access control and group administration. |
| Workbench Win | Licensed (free) | The Windows desktop application — same platform, easy install. A licence key is included at no cost. One version, no tiers. workbench-win.sentuny.com |
| Workbench Mac | Planned | The macOS desktop application — same easy-install approach as Workbench Win. Under consideration. As with all desktop versions, technically-inclined users may prefer the open-source Workbench directly. |
| Component | Requirement |
|---|---|
| Operating System | Linux (Ubuntu 20.04+, Debian 11+, or similar). Also runs on macOS for development. |
| Python | 3.10+ with FastAPI, SQLAlchemy, and database drivers |
| Database | PostgreSQL 12+ (required — stores Workbench metadata, user accounts, recipes, datasets) |
| Node.js | 18+ for the Next.js frontend |
| Memory | 2 GB RAM minimum, 4 GB recommended |
| Disk | ~1 GB for application + dependencies. Additional space for imported datasets. |
| Browser | Any modern browser — Chrome, Edge, Firefox, Safari |
| Target Databases (optional) | PostgreSQL 12+, MySQL 8+, SQL Server 2019+, or MS Access 2010+ via the MS Access Endpoint |