Migration · Data Console · Agent Gateway

Move your database.Govern every row.

Move schema and data across 11 engines with every row checksum-verified. Then browse, mask and audit that data — including when an AI agent is the one asking.

Start freeSee how it works

Free tier · no credit card · read-only credentials are enough to start

dbshift — CLIruns on your machine

# Read the source and convert its schema

$ dbshift analyze --source mysql://prod

# Move the data — batched, parallel, resumable

$ dbshift migrate --source mysql://prod --target postgresql://new

# Prove both sides hold the same rows

$ dbshift validate --source mysql://prod --target postgresql://new

Same engine support as the hosted app — nothing leaves your network.

0
Database engines
0
Certified paths
0%
Row-level verified
0
Downtime with CDC

Three products, one account

Everything you need to move, see and govern your data.

Migration

Move schema + data across 11 engines with conversion, validation, rollback, zero-downtime CDC and a guided go-live cutover — one-off, recurring on cron, or triggered from CI.

Start a migration

Data Console

Ask questions in plain English, then browse, query and govern any database — masked-on-read PII, access reviews, and masked test-data copies for staging. No SQL client.

Open the console

Agent Gateway

The safe data door for AI agents — scoped tokens, curated views, PII masking, and Slack-approved writes. MCP or REST, fully audited.

Open agent gateway

Built to be trusted with production data.

Schema conversion

Types, keys, indexes and constraints translated across every engine pair.

Checksum validation

Row-level, type-aware checksums prove every row landed — sampled or exact.

Zero-downtime CDC

Capture the change-log before the snapshot, replay after — no gap, no drift.

Data Console

Ask your data in plain English, then browse read-only — search, filter, follow foreign keys, export.

Guided cutover

Go-live as a staged flip: drain the change stream, verify every row, then switch. Abortable until the end.

Reconciliation monitors

Compare two databases on a schedule — get alerted the moment they drift apart after go-live.

Sensitive data

Find PII across all engines — emails, cards, SSNs — for compliance.

Masking policy library

Define your GDPR / PCI ruleset once, apply it to any migration in one click — seeded from the PII scan.

Agent usage budgets

Hard daily and monthly caps per agent token — an AI agent can never quietly page a table out.

Test data copies

Small, FK-intact, masked-by-default copies of production for dev and staging.

Recurring schedules

Cron-driven migrations and validations — skipped, never doubled, while a run is in flight.

Access reviews

One report of who can browse and who can unmask, across every data source. CSV for auditors.

Works with every major database

mysqlMySQL
postgresqlPostgreSQL
mariadbMariaDB
sqlserverSQL Server
oracleOracle
sqliteSQLite
mongodbMongoDB
cockroachdbCockroachDB
yugabytedbYugabyteDB
tidbTiDB
redshiftAmazon Redshift
mysqlMySQL
postgresqlPostgreSQL
mariadbMariaDB
sqlserverSQL Server
oracleOracle
sqliteSQLite
mongodbMongoDB
cockroachdbCockroachDB
yugabytedbYugabyteDB
tidbTiDB
redshiftAmazon Redshift
mysqlMySQL
postgresqlPostgreSQL
mariadbMariaDB
sqlserverSQL Server
oracleOracle
sqliteSQLite
mongodbMongoDB
cockroachdbCockroachDB
yugabytedbYugabyteDB
tidbTiDB
redshiftAmazon Redshift
mysqlMySQL
postgresqlPostgreSQL
mariadbMariaDB
sqlserverSQL Server
oracleOracle
sqliteSQLite
mongodbMongoDB
cockroachdbCockroachDB
yugabytedbYugabyteDB
tidbTiDB
redshiftAmazon Redshift
mysqlMySQL
postgresqlPostgreSQL
mariadbMariaDB
sqlserverSQL Server
oracleOracle
sqliteSQLite
mongodbMongoDB
cockroachdbCockroachDB
yugabytedbYugabyteDB
tidbTiDB
redshiftAmazon Redshift
mysqlMySQL
postgresqlPostgreSQL
mariadbMariaDB
sqlserverSQL Server
oracleOracle
sqliteSQLite
mongodbMongoDB
cockroachdbCockroachDB
yugabytedbYugabyteDB
tidbTiDB
redshiftAmazon Redshift
mysqlMySQL
postgresqlPostgreSQL
mariadbMariaDB
sqlserverSQL Server
oracleOracle
sqliteSQLite
mongodbMongoDB
cockroachdbCockroachDB
yugabytedbYugabyteDB
tidbTiDB
redshiftAmazon Redshift
mysqlMySQL
postgresqlPostgreSQL
mariadbMariaDB
sqlserverSQL Server
oracleOracle
sqliteSQLite
mongodbMongoDB
cockroachdbCockroachDB
yugabytedbYugabyteDB
tidbTiDB
redshiftAmazon Redshift

Connect once. Work your way.

01

Connect

Point at any database — any of 11 engines, in place, credentials never leave the platform.

02

Work

Migrate it, browse and query it, or hand an AI agent scoped, audited access — same connection, three paths.

03

Verify

Row counts, checksums and audit logs prove nothing was lost, leaked, or touched without a trace.

Working with AI and your data

What is actually AI here, what deliberately is not, and how agents get access without your database credentials.

How do I give an AI agent access to my database safely?

Register the database with Agent Gateway, set a policy — which tables and columns are readable, whether PII is masked, row caps, rate limits, usage budgets — and issue the agent a scoped token instead of your database credentials. The agent connects over MCP or REST and can only ever do what the policy allows. Every query is logged, and Replay shows exactly which rows the agent saw.

What is an MCP server for a database?

MCP (Model Context Protocol) is how AI clients like Claude Desktop and Cursor discover and call tools. Agent Gateway exposes your database as an MCP endpoint with structured tools — list tables, query a table, aggregate, follow a foreign key — rather than a SQL prompt. The agent never writes SQL, so column rules and PII masking hold on every result instead of depending on what the model chose to generate.

Can I let AI query production data without exposing customer PII?

Yes. Sensitive columns are found by a deterministic pattern scan and masked on read, so an agent or a teammate without an explicit grant gets redacted values. Masking is applied before the data leaves the server, and grouping or aggregating on a masked column is refused — otherwise 'group by email' would enumerate the very values the mask hides.

Can I ask my database questions in plain English?

Yes — Data Console's Ask answers questions like 'which customer spent the most?' from your real rows. You bring your own Anthropic or OpenAI key; DBShifts never supplies or bills for one. The model calls the same governed, read-only tools the browse UI uses, so it generates no SQL and never receives a masked value. Ask is off until you enable it per data source, and every question is written to the audit log.

Does DBShifts use AI to migrate, mask or validate data?

No, and that is deliberate. Schema conversion uses explicit type-mapping rules, PII detection uses pattern matching with checks like Luhn validation for card numbers, masked values are generated deterministically so joins and unique keys still work, and validation compares row counts and type-aware checksums. Those paths have to be repeatable and explainable, so no model is involved. AI is confined to the two places you opt into it: agents reading through the gateway, and Ask.

Work with your data, safely.

Start free — migrate a database, open the console, or connect your first agent in minutes.

Get startedSign in