Everything you get.
All of it, free.
From basic database sync to AI-powered pipelines, scheduled jobs, inline transforms, and a persistent knowledge engine โ Datavor ships all features across every version at no cost.
Sync any table with
one sentence.
The foundation of Datavor. Connect two databases, describe what you want, and the right data moves โ full, filtered, or incremental. No SQL, no config files, no pipelines to build.
Full Sync
Copy an entire table from source to target. Replaces all existing data. Best for first-time migrations or small tables.
Incremental Sync
Only sync rows changed since the last run using a timestamp column. Up to 98% faster on repeat syncs.
Partial Sync
Sync a filtered subset using any WHERE condition โ country, date range, status, or any column value.
โ 10,542 rows synced in 5.2s
"Sync only active US customers from the last 30 days"
โ 3,241 rows (WHERE country='US' AND active=1)
"Sync orders incrementally using updated_at"
โ 47 changed rows synced in 0.3s ยท 98% faster than full sync
Understand differences
before you sync.
Compare schemas, detect drift, and get AI-recommended sync order โ all before a single row moves.
Schema Comparison
Side-by-side diff of any table between two databases. Highlights missing columns, type mismatches, and index differences.
Schema Analysis
Full database diff โ every table, every column. Surfaces all differences before a migration begins.
Sync Order Recommendations
AI-powered suggestions for which tables to migrate first, based on size, dependencies, and foreign key relationships.
Browse your database
like a conversation.
Ask Claude to show you a tree of all tables, describe any table's structure, or run a query in plain English. No GUI needed.
Database Tree View
Hierarchical view of all tables, columns, and row counts. Perfect for databases with 50+ tables.
Table Inspector
Full column types, constraints, indexes, and foreign keys for any table โ in seconds.
Natural Language Queries
Describe what data you want, Claude writes and executes the SQL. Results come back clean and readable.
Set it once.
Run forever.
Create automated sync jobs using plain English schedules. The daemon runs independently of Claude Desktop โ your pipelines keep moving while you're away.
Natural Language Schedules
"Every night at 2am", "weekly on Sunday", "every hour" โ Datavor converts to the correct cron expression automatically.
Job Management
List, pause, resume, run manually, or delete any job. Full status including last run time and success rate.
Daemon Mode
Run node build/scheduler-daemon.js to keep jobs running 24/7 without Claude Desktop. pm2 compatible.
โ Job created ยท Daily 02:00 (0 2 * * *) ยท orders ยท full sync
Shape data
as it moves.
Rename columns, convert types, filter rows, remap values, and add computed columns โ all inline during sync. Zero extra passes, zero temp tables. Preview on real data before writing anything.
Rename
Rename any column during sync. customer_id โ cust_id
Cast
Convert column types. String score to float, datetime to timestamp, etc.
Filter
Drop rows that don't match. Keep only active = 1 rows.
Computed
Add derived columns. Combine first + last name into full_name.
Value Map
Replace values. A โ Active, I โ Inactive.
Preview
See before/after on 5 real sample rows before writing a single byte to the target.
Every sync recorded.
Always visible.
Datavor automatically records every operation to a local ledger. Ask Claude for a health check at any time โ success rates, row counts, failure diagnostics, and per-table history. Reads from a local file, instant, no database needed.
Summary View
Overall success rate, rows moved, most active tables, and a daily trend for any time window.
Table History
Full run log for any table โ every sync with row count, duration, mode, and incremental cursor.
Failure Diagnostics
Every failure logged with its error message and a specific fix suggestion โ not just a stack trace.
Five databases.
Any direction.
Datavor v2.0 connects to MySQL, PostgreSQL, SQL Server, SQLite, and Snowflake โ and syncs between any combination. One natural-language request, any source to any target.
Aiven
PlanetScale
Supabase
AWS RDS
Azure SQL
โ Types: DATETIMEโTIMESTAMP_NTZ, JSONโVARIANT, TINYINT(1)โBOOLEAN
"Copy production data from PostgreSQL to local SQLite for testing"
โ 45,231 rows ยท local dev environment refreshed
"Sync legacy SQL Server table to PostgreSQL with column renames"
โ Transforms applied inline ยท 22,000 rows in 8.4s
Datavor learns
your data over time.
The Context Engine builds a persistent local knowledge brain โ silently accumulating everything it learns about your databases, rules, and sync patterns.
Schema Memory
Remembers your schema across sessions. Automatically detects column additions, removals, and type changes.
Business Rules
Save rules that auto-apply to future syncs. "Always exclude @test.com emails" โ set once, applied forever.
Relationship Tracking
Foreign keys discovered from DB metadata and naming patterns. Sync pairs tracked across databases.
explain_database
Ask "Explain my production database" and get a plain-English summary of tables, rules, history, and stats.
Sync Intelligence
142 syncs recorded, 97.2% success rate, 890,000 rows moved โ all queryable through Claude.
100% Local
The context brain lives at on your machine. Schema metadata only โ never actual row data.
๐ production_db (MySQL)
23 tables known ยท 4 business rules (2 auto-applied)
142 syncs recorded ยท 97.2% success ยท 890k rows total
2 schema changes detected recently
First connected: 2026-01-15
๐ก Auto-rule active: exclude_test_orders
email NOT LIKE '%@test.com%' ยท applied 47 times
No surprises
between databases.
A canonical type system replaces all per-pair mappings. Every engine maps to and from a shared intermediate layer โ guaranteeing correct conversions across all five databases.
| Universal Type | |
|
|
SQLite |
Snowflake |
|---|---|---|---|---|---|
| boolean | TINYINT(1) | BOOLEAN | BIT | BOOLEAN | BOOLEAN |
| uuid | CHAR(36) | UUID | UNIQUEIDENTIFIER | TEXT | VARCHAR(36) |
| json | JSON | JSONB | NVARCHAR(MAX) | TEXT | VARIANT |
| datetime | DATETIME | TIMESTAMP | DATETIME2 | DATETIME | TIMESTAMP_NTZ |
| binary | BLOB โ ๏ธ | BYTEA โ ๏ธ | VARBINARY โ ๏ธ | BLOB โ ๏ธ | BINARY โ ๏ธ |
Real-time replication.
No polling.
Stream every INSERT, UPDATE, and DELETE from PostgreSQL (WAL) or MySQL (binlog) into your destination as it happens. Checkpoints are persisted in the Context Engine so streams resume cleanly after a restart.
PostgreSQL WAL
Logical replication via wal_level=logical. Slot position persisted across restarts.
MySQL Binlog
Row-based binary log streaming. Requires binlog-format=ROW. Binlog position persisted.
~50ms Lag
Changes arrive in the destination within milliseconds of being committed in the source. No batch windows, no stale reports.
โ CDC stream started.
Source: production_postgres.orders
Method: PostgreSQL logical replication (WAL)
Slot: datavor_orders_slot ยท LSN 0/16B3D80
Lag: ~50ms
"Show CDC status"
โ 1 active stream ยท 12,847 events applied ยท lag 48ms
Bad rows no longer
kill the pipeline.
One malformed row used to abort the whole batch. Now the bulk write tries first, falls back to per-row on failure, and isolates the offenders โ the rest of the batch commits successfully.
Bulk-first, row-fallback
Attempts the bulk write for speed. On failure, retries row-by-row to isolate exactly which rows are bad.
Failure Breakdown
Every sync reports Rows synced / Rows failed, broken down by error type: null_violation, duplicate_key, type_mismatchโฆ
Error Learner
Recurring error patterns are detected, counted, and surfaced as proactive suggestions โ before you notice the problem.
โ Synced 9,847 rows
โ Failed 3 rows
Failures by type:
null_violation: 2 (column 'email' is NOT NULL)
duplicate_key: 1
Run dashboard_failures for full row payloads.
Save transforms once.
Reuse forever.
Stop re-typing the same transform configuration on every sync. Save it as a named recipe, and apply it with one command across any environment. Frequently-used transforms are auto-captured from sync runs.
save_recipe
Save any transform configuration โ renames, casts, filters, computed columns โ as a named recipe.
list_recipes
Filter saved recipes by connection, table, or tags. Find the right one instantly across dozens of recipes.
apply_recipe
Load a recipe by name and apply it to any sync in a single command. One definition, zero drift across environments.
โ Recipe 'cleanup_orders' saved (recipe_id: 7)
30 days later:
"Apply the cleanup_orders recipe and sync orders to staging"
โ Recipe loaded ยท 12,420 rows synced ยท 0 failed
Datavor tells you what
to fix before you notice.
Schema changes, recurring errors, and slow syncs are detected, scored, and surfaced as actionable suggestions with accept/dismiss lifecycle. Dismissed suggestions are never re-raised.
Schema Changes
New column detected on a synced table? Datavor flags it immediately and proposes adding it to the destination โ never auto-runs DDL.
Error Patterns
Recurring failures are counted and surfaced with their error type and a suggested fix.
Optimisations
Full-replace syncs moving mostly-unchanged data are flagged with a switch-to-CDC or incremental suggestion.
3 suggestions pending:
1. [schema_change] New column 'discount_pct' on orders
Syncs to analytics.orders โ destination is missing it
2. [error_fix ยท 7ร] Null violation on shipments since Apr 12
3. [optimisation] events full-replace avg 84s โ switch to CDC?
"Accept 1 and 3. Dismiss 2 โ those NULLs are intentional"
โ Done. Suggestion 2 will not be raised again.
Wire real pipelines,
not timing guesses.
Express job dependencies as a DAG. Cycles are rejected at definition time. Failed parents block their children and retry with exponential backoff. No more "I set the rollup 30 minutes after the sync, it should be fine."
scheduler_add_dependency
Declare "job B runs after job A." Multiple parents supported โ the child waits for all of them.
scheduler_show_graph
Renders the full DAG topology โ parents, running children, waiting grandchildren, independent jobs.
Exponential Backoff
Failed jobs retry at 1m โ 2m โ 4m โ 8m intervals. Children are skipped until the parent succeeds.
โ 2 dependencies added. Cycle check passed.
orders_sync โ analytics_rollup
customers_sync โ analytics_rollup
"Show me the scheduler graph"
orders_sync [โ done]
customers_sync [โ done]
โโ analytics_rollup [โณ running]
โโ email_digest [โ waiting]
All databases.
All versions.
All five supported engines, plus cloud-hosted variants โ from the first release through v2.0.
Runs everywhere
you work.
Datavor v2.0 added full Windows and Linux support alongside the existing macOS release.
macOS
Apple Silicon and Intel. v1.0 and all subsequent versions. Homebrew-friendly.
Windows 10/11
Native Windows support added in v2.0. Works with Claude Desktop for Windows.
Linux
Ubuntu 22.04+ confirmed. v2.0 release. Docker-compatible for CI/CD environments.
Works with
any MCP client.
The same one-line config block works across Claude Desktop, Claude Code, Cursor, and Cline. If it supports MCP, it supports Datavor.
"mcpServers": {
"datavor": {
"command": "datavor"
}
}
}
Ready to start?
Install Datavor in 60 seconds. All features across every version, free forever.