Home Product Docs About ๐• @Datavor_ai โ–ถ YouTube
Get Started โ€” Free โ†’
โ† Blog
4 April 2026 ยท announcement ยท 7 min read

Datavor v2.0:
Five Databases. One Brain.

Datavor now connects to SQL Server, SQLite, and Snowflake โ€” and ships with the Context Engine, a persistent knowledge layer that silently learns your databases, rules, and patterns over time.

With v1.5, Datavor became a complete pipeline tool: schedules, transforms, a sync dashboard, and automatic type mapping between MySQL and PostgreSQL. It proved that natural language was a real interface for data engineering โ€” not a demo, but something you could actually rely on.

v2.0 goes further in two directions at once. Breadth: three new database connectors, so you can sync between any combination of five engines. Intelligence: the Context Engine, a persistent local knowledge store that turns Datavor from a tool you use into a tool that understands your data.

v2.0 General availability ยท 4 April 2026 ยท Free ยท 34 MCP tools
๐Ÿš€
Upgrading from v1.5? Run npm install -g datavor@latest and restart Claude Desktop. All existing connections, jobs, and sync history carry over automatically.

Five Databases, Zero Configuration

Datavor started with MySQL and PostgreSQL. v2.0 adds SQL Server, SQLite, and Snowflake โ€” so you can sync between any pair, in any direction. One tool. Five engines. All free.

๐Ÿ–ฅ๏ธ

SQL Server

โญ New in v2.0

Full sync, transforms, and schema introspection. Windows and Linux Docker tested. Handles NVARCHAR, DATETIME2, UNIQUEIDENTIFIER, and BIT natively.

๐Ÿ“„

SQLite

โญ New in v2.0

WAL mode for concurrency. UPSERT with AUTOINCREMENT detection. Perfect for local dev environments, edge deployments, and rapid data refresh.

โ„๏ธ

Snowflake

โญ New in v2.0

MERGE-based upsert, VARIANT for JSON, TIMESTAMP_NTZ for datetimes. Sync your operational MySQL or PostgreSQL data directly into your Snowflake warehouse.

Sync Between Any Two of Five Engines MySQL PostgreSQL SQL Server โญ new SQLite โญ new Snowflake โญ new โ—† Datavor any โ†’ any Describe what you want in plain English โ€” Datavor handles the rest
Figure 1 โ€” Five supported engines, any sync direction

The Context Engine: Datavor Gets Smarter Every Sync

Every ETL tool will run your pipeline. None of them remember it. Until now.

The Context Engine is a persistent local knowledge store that silently accumulates four kinds of understanding as you work โ€” schema snapshots, business rules, table relationships, and sync history. You don't set it up. You don't configure it. It just grows in the background, and when you need it, it's there.

Traditional tools move your data. The Context Engine understands it โ€” and gets better at it every single time.
The Context Engine โ€” A Brain That Grows With Every Sync ๐Ÿ”„ Sync runs rows ยท duration ยท errors ๐Ÿ—‚๏ธ Schema scans columns ยท types ยท changes ๐Ÿ“‹ Your rules "always filter test data" ๐Ÿ”— Relationships foreign keys ยท sync pairs ๐Ÿง  Context Engine context.db (local SQLite) schema snapshots business rules (auto-apply) table relationships sync history & error patterns silently learns โ€ข no setup required get_context "What do you know about orders?" โ†’ explain_database "Explain my production DB" auto-apply rules filters applied without asking sync #1 sync #47 sync #142 knowledge accumulates ยท switching cost grows ยท it just gets better
Figure 2 โ€” The Context Engine: four inputs, three outputs, zero configuration

What it learns automatically

Schema changes: every time you connect to a database, Datavor snapshots the column structure. If a column was added or removed since last time, it flags it โ€” before your next sync can break.

Business rules: tell Datavor once โ€” "always filter out test orders, emails that contain @test.com" โ€” and it saves that as a rule with auto_apply: true. Every future sync to that table applies the filter automatically, without you having to ask again.

Relationships: Datavor detects foreign keys from database metadata and column naming patterns, and tracks which databases you're syncing between. Ask it about a table and it already knows where the data is going.

Sync history: every run is logged โ€” rows moved, duration, errors, success rate. After 142 syncs, when you ask "how is my orders pipeline doing?", the answer is specific.

example โ€” get_context in action
"What do you know about the orders table?" โ†’ Claude responds with everything Datavor has learned: schema: 4 columns, last scanned 2026-03-20 recent changes: column "discount_code" added 2026-03-18 rules: "exclude_test_orders" (auto-apply, used 47 times) relationships: syncs to analytics_postgres.orders sync_stats: 142 syncs ยท 97.2% success rate

Universal Type Engine: O(n) Instead of O(nยฒ)

Every database has its own type names. BOOLEAN in PostgreSQL is TINYINT(1) in MySQL, BIT in SQL Server, and BOOLEAN in Snowflake โ€” which is different to how Snowflake handles VARCHAR. Multiply that across five engines and you get a combinatorial problem that's easy to get wrong.

v2.0 replaces the old per-pair approach with a canonical intermediate type layer. Every connector maps to and from one shared format. Adding a sixth database connector in the future requires one mapping file, not five.

Old: O(nยฒ) โ€” every pair needs a mapping 10 mapping pairs adding one more = +4 new pairs New: O(n) โ€” one canonical layer canonical type layer MySQL Postgres SQL Server SQLite Snowflake 5 mapping files adding one more = just +1 file
Figure 3 โ€” Universal type engine: hub-and-spoke replaces pairwise mappings

34 MCP Tools. One Natural Language Interface.

v2.0 brings the total to 34 MCP tools across seven categories. Five new tools in the Context Engine. Three new connection tools for SQL Server, SQLite, and Snowflake. Every tool works through the same interface: Claude Desktop, Cursor, or Cline.

34
MCP Tools total
5
Database connectors
5
Context Engine tools โญ
8
New tools in v2.0

Not Just Claude Desktop โ€” Works Everywhere MCP Does

v2.0 is the first version we've formally tested and confirmed across all three major MCP clients. The same config, the same tools, the same natural language experience โ€” wherever you code.

Claude Desktop

The original home for Datavor. Mac, Windows, and Linux all confirmed. Install in under 2 minutes.

Cursor

Add Datavor to .cursor/mcp.json in your project root. Sync databases from inside your editor.

Cline (VS Code)

Configure via Cline MCP Servers. Works with all 34 tools from within VS Code's Cline extension.

Where We're Headed Next

Core Sync Engine v1.0

MySQL โ†” PostgreSQL sync, full / partial / incremental modes, schema inspection. โœ“ Shipped.

Pipelines & Observability v1.5

Scheduler, Transform Pipeline, Sync Dashboard, Type Engine. โœ“ Shipped.

Multi-Connector + Context Engine v2.0

SQL Server, SQLite, Snowflake. Data Context Engine. Universal canonical type system. 34 MCP tools. โœ“ Shipping today.

CDC & dbt Integration v2.5

Change Data Capture for real-time sync. dbt integration for transformation workflows. Planned Q3 2026.

Visual Layer v3.0

Local web dashboard โ€” browse history, manage jobs, build transform pipelines visually. No more terminal required. Planned Q4 2026.

Get Started in Under 2 Minutes

Install via npm, add one config block to Claude Desktop (or Cursor or Cline), and you're connected to all 34 tools.

installation
# 1. Install or upgrade Datavor npm install -g datavor # 2. Add to Claude Desktop config # macOS: ~/Library/Application Support/Claude/claude_desktop_config.json # Windows: %APPDATA%\Claude\claude_desktop_config.json # Linux: ~/.config/Claude/claude_desktop_config.json { "mcpServers": { "datavor": { "command": "datavor" } } } # 3. Restart Claude Desktop โ€” look for the ๐Ÿ”จ icon
๐Ÿง 
Try the Context Engine first: connect to a database you use regularly, run a few syncs, then ask "Explain my production database" or "What do you know about the orders table?" โ€” and watch Datavor tell you everything it already learned.
โ—†
Datavor Team
Building AI-native database tools ยท @Datavor_ai
Share on ๐• Download v2.0 โ†’