All 100, in order.
-
Reason 1: Install it where you want it
DataMind OS runs on a mainstream Linux server with Docker (Ubuntu, Debian, RHEL, Rocky or Alma), in any cloud or your own data center, physical or virtual. Your lake, warehouse and dashboards stay on your servers. AI requests go through DataMind's AI gateway, which routes them only to zero-data-retention model endpoints.
Your servers, your control -
Reason 2: Live queries on ClickHouse
Lake and Warehouse data live in ClickHouse, the open-source columnar database built for real-time analytics, stored compressed with ZSTD. Every chart, export and subscription queries it live, with no extracts to schedule. Compression depends on your data, so DataMind shows each table's real compressed and uncompressed size.
Your servers, your control -
Reason 3: Everything you need, in one OS
Connectors, Lake, Warehouse, modeling, ETL, orchestration, lineage, BI, Slides, AI Minds, swarms, benchmarks, PII protection, governance, SQL editor and observability ship together. One product and one contract in place of a stack of separate tools.
Your servers, your control -
Reason 4: Guided connections, AI-built connectors
Pick a source, follow the wizard, test it, choose tables, or let the Lake Builder AI do it. Not listed? The connector AI reads the API docs or screenshots, calls the API and writes the connector with you.
Connect and load your data -
Reason 5: Designed down to the last detail
Smart alignment guides, keyboard move and resize with a live position badge, copy and paste of visuals and per-page Markdown notes on the report canvas, plus a sign-in screen whose characters follow your pointer and react as you type.
Built for the people using it -
Reason 6: Desktop and mobile app from day one
Every installation ships an installable DataMind OS app. Add it to your laptop or phone in a few taps, with no app-store rollout. Talk to your Minds on the go, with the full OS on desktop and tablet.
Built for the people using it -
Reason 7: Make it yours
Build color themes with your brand, chart and SQL-editor colors and background patterns, preview them live across the app, and, with the Publish Color Theme permission, share them in your team's gallery. Put your company logo and name on reports and decks.
Built for the people using it -
Reason 8: Eight load patterns, built in
Full reload, incremental reload, chunked backfill, full and incremental upsert, full and incremental snapshots that keep history, and one-time loads. Each is a setting, not a project.
Connect and load your data -
Reason 9: Access rules down to the row
On Lake tables and on warehouse dimensions, facts, dictionaries, static tables and views, choose which roles see the object, which columns and which rows. Rules compile into database views, so they hold in SQL, BI, AI and external tools.
Security and governance -
Reason 10: One sign-in across the apps
Sign in to the web and mobile app with your company SSO. Oda desktop, Oda for VS Code and Claude Desktop sign in through DataMind OS, so they follow the same SSO and 2FA. SQL and BI tools connect through each person's own synced database account, under the same access rules.
Security and governance -
Reason 11: An IDE-grade SQL editor in your browser
Autocomplete that knows your tables, columns, functions and 1,258 query settings, inline docs, 17 snippets, Alt+Enter rewrites, a formatter that refuses to change what your query does, live syntax checks and AI next-step suggestions.
Built for the people using it -
Reason 12: Talk instead of type
Voice input works in Minds chats and AI prompt boxes, in your chat's language, and costs no AI tokens. It works in Chrome, Edge, Mac Safari and Android. Audio is processed by your browser's speech service, not by DataMind.
Built for the people using it -
Reason 13: Work like on a desktop OS
Open tables, transformations and editors as floating windows, up to 20 each in Lake, Warehouse, Write back and BI. Place them side by side, minimize them to a tray and find them where you left them after a reload.
Built for the people using it -
Reason 14: Lineage that writes itself
Save a transformation, view or dictionary and DataMind asks the database engine which tables your SQL reads. No manual tagging. Your lineage map, from sources through Lake and Warehouse to models, Minds, decks and writebacks, stays current.
A warehouse that builds itself -
Reason 15: Skip the DDL and partition math
Deploy a table and DataMind writes its storage engine, partitioning and sort key. Lake tables partition from their load pattern, warehouse tables from locked system columns. Balanced, Turbo and Eco presets let you tune each load.
A warehouse that builds itself -
Reason 16: AI-written ETL, tested before it runs
AI writes or fixes the SQL and you review the diff. Columns map automatically, dependencies are detected and every load checks for duplicate keys. Test on real data without touching the target, then validate, review and finalize.
A warehouse that builds itself -
Reason 17: Nine load patterns for dimensions and facts
Full reload, diff upsert, insert, update and delete for dimensions; full, periodic and looped truncate-insert and periodic insert for facts. Plus automatic change history for dimensions. Pick a pattern and DataMind generates the load SQL.
A warehouse that builds itself -
Reason 18: Your warehouse workflow builds itself
DataMind already knows how your tables depend on each other. It orders every load, runs independent steps in parallel up to your batch size, makes dependent steps wait and refuses circular dependencies.
A warehouse that builds itself -
Reason 19: Retire your mapping spreadsheets
Static tables and dictionaries keep reference data and mappings inside DataMind: type them, upload Excel or let AI fill them. Dictionaries refresh from sources on schedule and count unmapped values. A SQL alert can email you when new ones arrive.
A warehouse that builds itself -
Reason 20: Find any table in seconds
Press Alt+Q on any screen to search every Lake table, Warehouse object and transformation you can access, and open it in a window right where you are.
Built for the people using it -
Reason 21: ER diagrams built in
Import from a relational database or Snowflake and DataMind reads its foreign keys and draws the diagram. Drag between fields to add relationships, auto-arrange in one click, or ask the Lake Builder AI to suggest undeclared keys.
A warehouse that builds itself -
Reason 22: Watch workflows live, then rewind
See every run as a live diagram with timing, rows and errors for each step. Drag the timeline to any moment to replay exactly what was running, and open any previous run the same way.
A warehouse that builds itself -
Reason 23: Zero-data-retention routing on every model call
DataMind's AI gateway sends model and embedding requests only to zero-data-retention provider endpoints, denies provider data collection and disables any model that loses its ZDR endpoint. DataMind's platform keeps its own request log; turn on Prohibit PII transmission to keep request bodies out of it.
Private AI and Kartlos -
Reason 24: Pick speed or price for catalog models
Pay-as-you-go catalog models run on several zero-data-retention provider endpoints, with uptime checked every 30 minutes. Your company chooses whether providers are ordered by throughput, latency, price or a balance of all three.
Private AI and Kartlos -
Reason 25: Leading AI models in one place
Next to the Kartlos tiers, choose from a curated catalog of frontier models, all on zero-data-retention routing. DataMind adds new models as ZDR endpoints appear. Admins choose which are allowed, and each Mind picks from that list.
Private AI and Kartlos -
Reason 26: EU-region AI routing
One company setting sends your AI chat traffic to EU-region endpoints, and EU is the default for new companies. EU mode offers the catalog models served in the EU, with Kartlos set to EU endpoints.
Private AI and Kartlos -
Reason 27: Kartlos model tiers, built in
Kartlos Premium and Kartlos Standard are DataMind's built-in model tiers, backed by curated frontier models: Premium for complex reasoning and agentic work, Standard for quick everyday tasks. Every month, 85M Premium and 850M Standard tokens are included in your license, and DataMind's own agent instructions never count against them. Each web search uses 1,500 Kartlos Premium tokens and each page read 2,000.
Private AI and Kartlos -
Reason 28: Built-in MCP server
Connect Claude Desktop to DataMind OS with one downloadable extension. Each person signs in with their own DataMind account, and Claude can use only the Minds and data that person may access, read-only.
AI Minds and agents -
Reason 29: SSO in a guided wizard
Connect Microsoft Entra ID, Google Workspace or any OpenID Connect provider in a four-step wizard with a test sign-in. Colleagues from approved domains get an account with your chosen default role on first sign-in.
Security and governance -
Reason 30: Built-in two-factor authentication
Protect sign-in with any authenticator app. Require 2FA for everyone or for specific users, enroll by QR code, recover with one-time codes, and let admins reset it when a phone is lost.
Security and governance -
Reason 31: Built-in notifications
Send workflow failure alerts to Email, Slack, Discord or Google Chat. Schedule BI reports, Slides decks and data alerts by email, with each BI report rendered under the recipient's row-level security. Add your mail server or webhook once.
BI, Slides and delivery -
Reason 32: AI analysts on your own data
Create Minds that answer business questions by writing and running SQL on the Lake and Warehouse tables you link. By default, every query runs on your own ClickHouse under the Mind's own database account.
AI Minds and agents -
Reason 33: The AI sees tokens, people see names
Columns marked as personal data reach the model only as vault tokens, while people in the chat see the real values. The same protection covers Oda and MCP queries.
Private AI and Kartlos -
Reason 34: Swarms: many agents, one answer
Turn on Swarm and a Mind splits a large task across parallel sub-agents, up to 5 per agent and 3 levels deep. Each starts with its own brief and the results it needs, then reports back.
AI Minds and agents -
Reason 35: Benchmark AI on your own questions
Build a multiple-choice question bank for a Mind and score the model, thinking-effort and response-mode combinations you pick, up to 5,000 items per experiment. Roll out the settings that get your questions right.
AI Minds and agents -
Reason 36: An AI data engineer for ingestion
The Lake Builder explores sources, previews samples with personal data excluded, infers keys, imports tables, sets load patterns, deploys, builds workflows and monitors loads. In Manual mode, every change waits for your approval.
AI Minds and agents -
Reason 37: An AI data engineer for your warehouse
The Warehouse Builder builds semantic sources, facts, dimensions, views, dictionaries and transformations through chat, using the same deploy, test and finalize steps as your team, and makes changes under your Warehouse permissions. You choose Manual, Auto-accept or Unrestricted approval. Available to admins.
AI Minds and agents -
Reason 38: Oda, an AI coworker on your desktop
On Windows, Oda signs in with your DataMind OS account, plans and carries out tasks, creates Word, Excel and PowerPoint files, researches the web and uses a browser you share. Company data comes only through your Minds, read-only.
AI Minds and agents -
Reason 39: From question to deck in one chat
A Mind turns its findings into a slide deck right in the conversation. Preview it, present it fullscreen, download PPTX or PDF, or open it in Slides to keep editing.
BI, Slides and delivery -
Reason 40: Decks that update themselves
Bind slide charts, tables and text to SQL on your Lake or Warehouse data. They refresh when the deck opens or presenting starts, and keep the last result if a refresh fails. Binding SQL needs its own permission.
BI, Slides and delivery -
Reason 41: An AI that builds and edits decks
Slides AI reads your deck and adds, edits, reorders and removes slides and elements, with web search when it needs facts. Every AI step undoes with Ctrl+Z, and new slides come with motion already designed.
BI, Slides and delivery -
Reason 42: Show a picture, get a visual
Describe a chart or attach a screenshot, and the Custom Visual AI writes the code, bound to your data. Watch the tile change live on the canvas while the AI renders previews to check its own work.
BI, Slides and delivery -
Reason 43: Compare two periods in one visual
Matrix, table, KPI, bar, line, pie, funnel and key tree visuals compare two slicer scopes side by side, such as this year against last, with difference and percentage change from one server-side query.
BI, Slides and delivery -
Reason 44: Maps down to individual buildings
The built-in Map visual draws areas, points, heatmaps, 3D columns and buildings from country, region, city or coordinate fields, with 3D terrain, a globe view and labels in 100 languages, on DataMind's own map service.
BI, Slides and delivery -
Reason 45: Real PowerPoint files, not pictures
Slides decks export to PowerPoint with native charts and an embedded workbook, so Edit Data works, and with transitions including Morph. PDF, notes pages and slide images are one click away too.
BI, Slides and delivery -
Reason 46: A privacy model inside your stack
A multilingual PII detection model runs on your own servers and checks what people send to Minds, from chat messages and edits to report and task prompts, before it reaches a model. Each policy blocks, warns or tokenizes, with regex and custom rules alongside.
Private AI and Kartlos -
Reason 47: 22 built-in sources, ready to connect
SQL Server, Postgres, Oracle, MySQL, MariaDB, ClickHouse, Snowflake, Redshift, BigQuery, Azure SQL, MongoDB, Firestore, Firebase Realtime Database, RabbitMQ, webhooks, S3, Athena and five file formats, plus two build-your-own options.
Connect and load your data -
Reason 48: Big files, loaded without scripts
Load CSV, TSV, JSON and Parquet files up to 50 GB and Excel workbooks up to 1 GB, by resumable upload, from SFTP, FTPS or FTP servers, or from S3-compatible buckets. CSV, TSV and JSON files can arrive .gz or .zst compressed, and zip archives work too.
Connect and load your data -
Reason 49: Drop a spreadsheet into the conversation
Drop up to five Excel or CSV files into a Mind chat. Each sheet becomes a table the Mind queries with SQL next to your warehouse data, visible only in that chat, with an optional personal-data check first.
AI Minds and agents -
Reason 50: Install and run it from a browser
The DataMind Installer sets up DataMind OS on one Linux host in about 20 minutes on the recommended hardware, with a demo semantic model and report to explore. Its web console, with its own users, roles and audit trail, handles start, stop, restart, configuration with encrypted secrets, TLS settings and updates with live logs. The 2-month free trial runs on your own server.
Your servers, your control -
Reason 51: Operate the platform from inside it
Admin > Platform shows every service's state. Authorized admins restart or stop services, update the whole stack or selected services, follow live jobs and search container logs, all without opening SSH.
Your servers, your control -
Reason 52: Teach your AI company knowledge
Domain experts write markdown training cards that a Mind always applies or retrieves by similarity search. Upload PDFs and manuals as cards, start from 14 behavior templates and keep a version history of every card.
AI Minds and agents -
Reason 53: Least-privilege AI, column by column
Pick exactly which columns of each linked table a Mind may read. DataMind grants only those columns to the Mind's own database account, and you choose structure only, sample values or full data.
Private AI and Kartlos -
Reason 54: One Mind, each person's own rows
Switch a Mind to Execute as User and every query runs under the asking person's own database account, so their access rules and row-level security apply. One Mind can serve a whole department.
Security and governance -
Reason 55: AI queries with hard limits
By default, an analytical Mind queries through its own read-only database account, which cannot change structure or lift its own limits: 100 seconds and 4 GiB per query by default, set per Mind. In Execute as User mode, the asking person's own account and limits apply.
Private AI and Kartlos -
Reason 56: 20+ specialist agents across the OS
More than 20 purpose-built AI agents, from column commenter and key finder to measure author, slides author and semantic-model AI, share more than 250 platform tools. AI is built into the modules, not bolted on as one chatbot.
AI Minds and agents -
Reason 57: An AI co-author for semantic models
Model AI edits your semantic model live, profiles data, writes measures and relationships, runs tests and health checks, and logs every change. Revert a whole turn or a single change in one click.
BI, Slides and delivery -
Reason 58: Build the chart your team imagines
A sandboxed custom visual studio with a code editor, live preview, diagnostics and history, plus 12 bundled libraries including d3, ECharts, Vega and three.js. Visuals filter other tiles and export to PPTX and PDF.
BI, Slides and delivery -
Reason 59: A company-wide PII safety net
Turn on Prohibit PII transmission and DataMind's AI gateway scans every request, including tool arguments, embeddings and web searches, and refuses any that carry common personal identifiers. Refused requests use no tokens and are not stored.
Private AI and Kartlos -
Reason 60: Oda inside VS Code and Cursor
The Oda extension for VS Code and Cursor on Windows brings a coding agent connected to your company data into the editor, with optional inline suggestions, next-edit suggestions and change-by-change review.
AI Minds and agents -
Reason 61: See and fix what the database is doing
The ClickHouse console shows running queries, merges, mutations, memory, per-table storage, query history and a DDL log. Terminate a query, kill a mutation or optimize a fragmented table without a separate DBA tool.
Your servers, your control -
Reason 62: Custom connectors, coded by AI
Need full control? Connect Anything creates a Python connector from a template on the open-source Singer SDK, with an agent that calls the real API, writes the code and tests it. Secrets stay in a write-only vault.
Connect and load your data -
Reason 63: Feed APIs from your own tables
Grant an API recipe one Lake or Warehouse object and it can make a call per row, for example per customer ID. Personal-data columns go out as tokens, and every grant is confirmed by name and revocable.
Connect and load your data -
Reason 64: AI-written connectors stay on a leash
API recipes run inside the database engine with signed run instructions, call only approved hosts, refuse private-network addresses, verify TLS by default and have their credentials masked in database logs.
Security and governance -
Reason 65: Every chart is an API
Each chart returns CSV, JSON or Parquet through an API secured by personal or chart-scoped tokens and row-level security, with ready snippets for Python, JavaScript, C#, Excel Power Query and Google Sheets.
BI, Slides and delivery -
Reason 66: Measures in plain SQL that rename safely
Write measures in ClickHouse SQL that reference other measures and window functions. Dependencies come from the engine, and renaming a measure rewrites and re-validates every dependent one in a single transaction.
BI, Slides and delivery -
Reason 67: 130 measure recipes, no SQL required
The measure builder offers 130 ready-made recipes in 10 categories, from share of total and time intelligence to running windows, ranking and business KPIs. Pick columns and it writes the SQL.
BI, Slides and delivery -
Reason 68: Regression tests for your metrics
Attach up to 50 tests to each measure and run them against live warehouse data from the editor, the Health tab or the Model AI. Catch a source change that breaks revenue before the board report goes out.
BI, Slides and delivery -
Reason 69: One governed business layer for every report
Semantic models turn Lake or Warehouse tables into BI tables with relationships, a diagram, docs, health checks, row-level security and a change history, plus a list of every report built on the model.
BI, Slides and delivery -
Reason 70: Know what breaks before you delete
Every Lake and Warehouse table lists the Minds, decks, semantic models, views and transformations that use it. Delete a table and DataMind shows those dependents first and asks you to confirm.
A warehouse that builds itself -
Reason 71: Pipelines that wait for each other
When a transformation reads a table loaded by another workflow, DataMind adds a gate that waits for that workflow's run of the same date and fails with a clear reason on timeout. No fragile time-based scheduling.
A warehouse that builds itself -
Reason 72: Restarts that resume, retries that heal
Restart a failed run and only unfinished steps execute again. Dependents of a failed step are skipped with a reason, and up to 10 automatic retries handle temporary memory, timeout or connection errors.
A warehouse that builds itself -
Reason 73: Cancel a load, lose nothing
Loads land in a temporary table and swap in whole or by day, so reports never read half-loaded data. Stop a sync midway and DataMind restores the table to its state before the sync started.
Connect and load your data -
Reason 74: Schema drift, handled safely
Sync Schema finds new columns and tables without loading them until you opt in, flags columns whose source type changed, and refuses any type change that would turn stored values into NULL.
Connect and load your data -
Reason 75: Configure hundreds of tables at once
Batch Edit opens every table of a datasource, or any warehouse scope, in one grid. Change sync types, incremental windows, schedules, permission groups and query settings for all selected or filtered rows in one save.
Connect and load your data -
Reason 76: Webhooks straight into your Lake
Get a permanent endpoint that stores every pushed event as a row, with sender presets, HMAC signatures, replay-window protection and secret rotation that accepts old and new secrets. Reload tables from the retained events.
Connect and load your data -
Reason 77: Message queues, buffered for safe reloads
RabbitMQ sources declare a queue per table or consume existing queues, discover them through the management API and keep consumed messages for 30 days by default, so a failed load can simply run again.
Connect and load your data -
Reason 78: Decade-old databases connect as they are
One SQL wizard covers SQL Server, Postgres, MySQL, MariaDB, ClickHouse, Oracle, Redshift and Azure SQL, with drivers for Oracle 11g and SQL Server 2005 or older, TLS, and Entra service-principal sign-in for Azure SQL.
Connect and load your data -
Reason 79: Messy files load without hand-cleaning
Delimiter, header and encoding are detected for you, extra rows at the top or bottom can be skipped, and date and decimal styles are handled. Rescue mode loads bad values as NULL and keeps the raw text instead of failing.
Connect and load your data -
Reason 80: Years of daily files, one table
Group files by pattern into one table, take the date from each file name for incremental loads, track every file's status, flag duplicates and reload only the affected days when late files arrive.
Connect and load your data -
Reason 81: Timestamps that match the source
Each connector keeps its own source timezone. The connection test detects it, daylight-saving days use that date's real offset, and out-of-range or unparseable dates become NULL or are clamped instead of failing the load.
Connect and load your data -
Reason 82: See the query behind every AI step
Each step a Mind takes appears as a card with its SQL and result, a live plan shows progress, and Tool History lists every step, so you can check where each number came from.
AI Minds and agents -
Reason 83: AI that wakes up when a load fails
Schedule a Mind prompt daily, weekly or monthly, or trigger it when a watched Lake or Warehouse workflow fails or finishes. Each task keeps run history and restorable versions.
AI Minds and agents -
Reason 84: See AI adoption and quality at a glance
Mind Analytics shows tokens, messages and conversations per Mind, user and role, plus lists of heavy chats, slow responses, artifact errors, likes and dislikes.
AI Minds and agents -
Reason 85: An AI audit trail on your own server
AI requests from Minds, agents, embeddings, web search and Oda are logged in your own database with model, agent, user, duration, status and request and response bodies, and export as CSV, Excel, JSON or TSV. Admins can let chosen people turn off Oda content logging; those requests keep usage and cost only.
Security and governance -
Reason 86: Permission presets instead of a matrix
Role permissions cover 12 areas, from Data Layer and Minds to DataPrism, Slides and Platform, each split into high-level, standard and granular groups. Any user can also get custom permissions directly.
Security and governance -
Reason 87: Answer the auditor in one click
One Audit button downloads a seven-sheet Excel workbook of users, roles, memberships, role permissions, each user's effective permissions and the permission catalog, stamped with who generated it and when.
Security and governance -
Reason 88: See what a user sees, fully audited
Admins with the Impersonate User permission see DataMind OS exactly as another user does, with no shared passwords. Every request is logged with both the viewed user and the real person behind it.
Security and governance -
Reason 89: SSO-only sign-in with a safe fallback
Require everyone to sign in to DataMind OS through your identity provider, while built-in admins keep a password fallback. Switching it on ends every existing session.
Security and governance -
Reason 90: Secrets that are never shown again
Connector, SSO, 2FA and MCP secrets are encrypted at rest with AES-256-GCM. Saved passwords are never shown again, and an optional policy forces re-entry when a connector's host or username changes.
Security and governance -
Reason 91: 53 personal-data types, tokenized for restricted roles
Tag Lake and Warehouse columns with one of 53 personal-data domains, from identity and contact to financial, health and credentials. Restricted roles see vault tokens that still work for joins and analysis.
Private AI and Kartlos -
Reason 92: 15 visual types, from KPI to key tree
Bar, combo, line, area, pie, donut, funnel, map, KPI, cards, table, matrix with drill-down, key tree, slicers and custom visuals, plus conditional formatting with data bars and icon sets.
BI, Slides and delivery -
Reason 93: Every report save is a version
Each report keeps its last 100 versions, with undo and redo, conflict handling, a live preview of any version, a diff viewer and one-click restore. A bad edit is one restore away.
BI, Slides and delivery -
Reason 94: Lineage that stays readable at scale
Switch between Overview, Detailed and Focus views, follow one object upstream and downstream to the depth you choose, search tables and columns and export PNG. Each viewer sees only what they may access.
A warehouse that builds itself -
Reason 95: Backfills by setting, not by script
Periodic loads expose synchronize days, loop increments, a full-reload start date and full reload when a partition is empty. Any run or test can override days and increments, so backfills need no one-off code.
A warehouse that builds itself -
Reason 96: Reload one bad day without a DBA
Truncate selected partitions by source system and date, drop stored data while keeping the definition, and see every run with rows, duration and errors. Abort a running load from the same screen.
A warehouse that builds itself -
Reason 97: From result to chart, export or alert
The SQL editor adds typed parameters, console history with side-by-side diff, server-side charts and pivots over the whole result, full exports to Excel, CSV, JSON or TSV, and an alert built from any query.
Built for the people using it -
Reason 98: Live health for every service
System Overview charts each service's memory, CPU and storage with incident detection. Live Metrics refreshes every 5 seconds and shows up to 90 days of uptime from 5-minute health checks.
Your servers, your control -
Reason 99: An Installer that rolls itself back
The DataMind Installer updates itself while DataMind OS keeps running, waits up to 300 seconds for the new build to report healthy, and rolls its own image back if it does not.
Your servers, your control -
Reason 100: Help where the question comes up
An Ask AI dock on every page answers product questions from the DataMind documentation, cites its sources, knows which page you are on and keeps the conversation while minimized.
Built for the people using it