Datavor is a complete data pipeline platform — sync, real-time replication, scheduling, intelligence, and visibility — all running locally, all driven by natural language through any MCP-compatible AI tool.
Full, incremental, and partial syncs between any two databases. Append, upsert, and merge modes. Inline transforms with saved recipes. Per-record fault tolerance — a bad row never aborts the whole job.
Sub-second change-data-capture via Postgres logical WAL and MySQL binlog. Tail your transaction logs, replicate INSERTs / UPDATEs / DELETEs to any target in real time. Live monitor at localhost:3000/cdc shows lag and event throughput.
Cron-style scheduled jobs with explicit dependency graphs. Jobs wait for upstream jobs to finish. Pause, resume, run-now, dependency visualization. Replaces Airflow for the 80% of teams who don't need a separate orchestration platform.
A local React dashboard at localhost:3000 showing live sync activity, connections, CDC streams, scheduler DAGs, and pending suggestions. Read-only by design — Claude still drives writes, the UI just makes inspection effortless.
A persistent local SQLite store of everything Datavor learns about your data — relationships, business rules, transform recipes, error patterns, accepted suggestions. The reason Claude becomes more useful with Datavor over time, not less.
~/.datavor/context.dbFrom the moment data lands in a source database to the moment it powers a dashboard or AI agent — Datavor covers every step.
Five database engines, one MCP server. Postgres, MySQL, SQL Server, SQLite, Snowflake — all through natural language.
Sync, replicate, or stream. Batch jobs for warehouses, CDC for real-time. Per-record fault tolerance throughout.
Inline or saved as recipes. The Context Engine remembers your transforms and applies them to future syncs automatically.
Schedule with DAGs. Monitor through the Web UI. Get proactive suggestions when schemas drift or syncs underperform.
Other data platforms try to be friendlier with SQL editors and drag-and-drop UIs. Datavor is friendlier because it speaks the language you already speak — and lives where your AI tools already live.
No DSL, no YAML, no drag-and-drop. Just describe what you want — "sync orders incrementally to analytics every hour, depending on customers" — and your AI tool handles the rest through Datavor's 47 MCP tools.
Datavor runs entirely on your machine. Your databases, credentials, and pipelines stay where they are. No vendor lock-in, no data egress fees, no shared infrastructure. Free tier is 100% local — no telemetry, no phone-home.
The Context Engine — Datavor's persistent local knowledge store — accumulates everything your AI learns about your data. Rules, recipes, relationships, error patterns. Every session starts smarter than the last.
No download. No signup. No card. Just npx datavor from any MCP-compatible AI tool.