Everything your databases need.
Nothing you don't.
See differences before you sync.
Ask Claude to compare any table between your MySQL and PostgreSQL databases. Get a clean visual diff showing missing columns, type mismatches, and index differences — all before a single row is touched.
- Side-by-side column comparison
- Type mismatch highlighting
- Complete migration analysis across all tables
- AI-powered recommendations on what to fix first
Stop re-syncing everything, every time.
After your first full sync, Datavor remembers the checkpoint. Every subsequent sync fetches only new and changed rows — making daily updates 50–98% faster as your tables grow.
- Timestamp-based change detection
- Automatic checkpoint tracking
- Works with any table that has an updated_at column
- Falls back to full sync gracefully
after first run
Five databases. One tool. Any direction.
Datavor v2.0 connects to MySQL, PostgreSQL, SQL Server, SQLite, and Snowflake — and syncs between any combination. One natural-language request, any source to any target.
- All 5 engines fully supported — read, write, sync, transform
- Universal type engine with O(n) complexity — no per-pair mappings
- Cross-engine migrations: MySQL → Snowflake, SQLite → PostgreSQL and more
- 34 MCP tools across all connectors
✅ Connected MySQL · Snowflake
✅ DATETIME→TIMESTAMP_NTZ, JSON→VARIANT
✅ 45,231 rows synced in 18.2s
Set it and forget it.
Create automated sync jobs in plain English. The daemon runs independently — your pipelines keep moving while you sleep.
- Create jobs in seconds — "sync orders every night at 2am"
- Natural language to cron: daily, hourly, weekly, monthly
- Pause, resume, or run any job manually on demand
- pm2-compatible daemon for 24/7 uptime
Shape data as it moves.
Rename columns, convert types, filter rows, remap values, and add computed columns — all inline during sync. Preview on real data before writing anything.
- Rename: customer_id → cust_id
- Cast: string score → float · Filter: active = 1 only
- Computed: first + last → full_name
- Preview on 5 sample rows before committing
Every sync recorded. Always visible.
Datavor automatically logs every operation. Ask Claude for a health check — success rates, row counts, failure diagnostics, and table history.
- Summary: success rate, rows moved, daily trend
- Table history: every run with row count and duration
- Failure diagnostics: error message + specific fix hints
- Zero config — all data written automatically
Real-time replication. No polling.
Stream every INSERT, UPDATE, and DELETE from PostgreSQL (WAL) or MySQL (binlog) into your destination as it happens. Checkpoints survive restarts.
- PostgreSQL: logical replication (wal_level=logical)
- MySQL: binary log streaming (binlog-format=ROW)
- Sub-100ms lag · LSN / binlog position persisted
- Prerequisites checked automatically before starting
Bad rows no longer kill the pipeline.
One malformed row used to abort the whole batch. Now the bulk write tries first, falls back per-row on failure, and isolates offenders — the rest commits.
- Reports Rows synced / Rows failed on every run
- Failures captured by type: null_violation, duplicate_key…
- Full failed-row payloads available via dashboard_failures
- Error Learner tracks patterns and surfaces fix suggestions
Datavor learns your data over time.
A persistent local knowledge brain silently accumulates everything it learns — schema changes, business rules, sync patterns, and relationships.
- Remembers schema and detects changes automatically
- Learns business rules — auto-applies them on future syncs
- Tracks foreign key relationships across databases
- "Explain my production database" → full plain-English briefing
Save transforms once. Reuse forever.
Stop re-typing the same transform configuration on every sync. Name it, save it, and apply it with one command across any environment.
- Save any transform set as a named recipe
- Apply with one command: "apply the cleanup_orders recipe"
- Filter by connection, table, or tags with list_recipes
- Frequently-used transforms auto-captured from sync runs
Datavor tells you what to fix before you notice.
Schema changes, recurring errors, and slow syncs are detected, scored, and surfaced as actionable cards. Accept or dismiss with one command.
- Schema changes: new column detected on a synced table
- Error patterns: recurring failures counted and explained
- Optimisation: full-replace syncs that should be CDC or incremental
- Dismissed suggestions never re-raised; accepted ones return action_config
Wire real pipelines, not timing guesses.
Express job dependencies as a DAG. Cycles are rejected at definition time. Failed parents block children; retries use exponential backoff.
- scheduler_add_dependency: "B runs after A"
- scheduler_show_graph: renders the full topology
- Cycle detection at definition time — invalid DAGs rejected
- Failed jobs retry 1m → 2m → 4m → 8m backoff
Browse your database like a conversation.
Ask Claude to show a tree of all tables, describe any schema, or run a query in plain English. No GUI needed.
- Database tree view — all tables, columns, row counts
- describe_table: types, constraints, foreign keys
- Natural language queries — Claude writes and runs the SQL
- compare_table_schemas — side-by-side diff before migration
📂 production_mysql (23 tables)
├─ orders (142,380 rows)
├─ customers (89,241 rows)
├─ products (4,521 rows)
└─ ... 20 more tables
"Describe the orders table"
✅ 8 columns · 2 indexes · 2 foreign keys detected
Fast. Free.
Built to scale.
Works wherever
Claude works.
One line of config. Any MCP-compatible client. Claude Desktop, Claude Code, Cursor, Cline — Datavor works with all of them.
Datavor MCP
MySQL·
PostgreSQL·
SQL Server·
SQLite·
Snowflake
What teams actually
use Datavor for.
Nightly Production → Analytics Sync
Mirror production MySQL to analytics PostgreSQL every night. Zero manual work.
↓ saves ~3h/weekReal-Time CDC Pipeline
Stream every INSERT, UPDATE, DELETE into analytics within 50ms. No polling, no stale reports.
~50ms lagData Migration with Transforms
Move tables between engines with inline renames, casts, and value remaps. Save as a recipe for future runs.
reusable recipesAI-Powered Pipeline Management
Datavor learns your schema, auto-applies your rules, spots errors, and suggests optimisations before you notice.
self-managing pipelinesCommon questions.
Is Datavor really free?
Yes — completely free in Phase 1. No credit card, no time limit, no feature restrictions. All 45 MCP tools included.
What databases does Datavor support?
MySQL, PostgreSQL, SQL Server, SQLite, and Snowflake — plus cloud variants: Supabase, AWS RDS, Azure SQL, Aiven, and PlanetScale.
What is CDC and do I need it?
Change Data Capture streams every row-level change in real time using PostgreSQL WAL or MySQL binlog. You need it if analytics data needs to be current within seconds. Regular syncs are fine for hourly/daily freshness.
Does Datavor work without Claude Desktop open?
Yes — run node build/scheduler-daemon.js (or use pm2) and all scheduled jobs and CDC streams run 24/7 without Claude.
What MCP clients work with Datavor?
Claude Desktop, Claude Code, Cursor, and Cline — all confirmed. Any MCP-compatible client works with the same one-line config.
Is my data sent to any server?
No. Datavor runs entirely on your machine. Your actual row data never leaves your environment.
What is the Context Engine?
A persistent local knowledge brain that learns from every sync — saving business rules that auto-apply, tracking history, and proactively surfacing suggestions when something looks off.
How do I get started?
Run npm install -g datavor, add one line to your Claude Desktop config, restart Claude, and ask "connect to my database at..." — syncing in under 5 minutes.
One command.
Infinite possibilities.
Install Datavor, add one line to your Claude Desktop config, and start syncing databases through natural language in under 5 minutes.