02 / MANAGED DATABASES

Dedicated NVMe PostgreSQL, MySQL & Redis.
Branch production data in under a second.

Full-speed local NVMe storage, continuous point-in-time recovery, optional high-availability replicas, and copy-on-write database branching built for pull requests and autonomous AI agents.

shell — provision Postgres 17 & create a copy-on-write PR branch
NVME + COW BRANCH
# Provision dedicated NVMe PostgreSQL 17 in eu-central
$ uncage workloads create \
    --project proj_prod \
    --name pg-prod \
    --surface database \
    --image postgres:17 \
    --vcpu 4 --mem 8192 --port 5432

✔ pg-prod running   (4 vCPU · 8 GB · 40 GB NVMe · eu-central)
  Internal URI: postgres://app:••••@pg-prod.proj-prod.internal:5432/app

# Fork an isolated copy-on-write branch for a schema migration test
$ uncage workloads fork pg-prod --name pg-prod-pr-418
✔ pg-prod-pr-418 ready in 640ms  (zero data copy, isolated writes)
Regions: us-east · eu-central · us-west · ap-south Launch Postgres 17 →
DB Branch Clone
< 1.0s
Storage Media
Dedicated NVMe
Engines
PG 17 · MySQL 8.4 · Redis 7
Disk Rate
$0.10/GB-mo
MANAGED ENGINES

Three battle-tested engines on dedicated NVMe.

RELATIONAL + PGVECTOR PORT 5432

PostgreSQL 17

  • ✓ Copy-on-write branching (<1s)
  • ✓ Continuous WAL PITR
  • ✓ pgvector & PostGIS ready
  • ✓ Optional HA read/failover replicas
RELATIONAL OLTP PORT 3306

MySQL 8.4 LTS

  • ✓ Dedicated local NVMe InnoDB
  • ✓ Continuous binlog archiving
  • ✓ Instant PR branch clones
  • ✓ Private project DNS + TLS
IN-MEMORY CACHE & QUEUES PORT 6379

Redis 7

  • ✓ Sub-millisecond internal latency
  • ✓ AOF + RDB durable snapshots
  • ✓ Dedicated RAM with zero noisy neighbors
  • ✓ Co-located with apps & agents
DATABASE ENGINE SPECS

Built for low-latency IOPS and fearless schema changes.

DB-01 Local IOPS

Dedicated NVMe Storage

Databases run on dedicated high-IOPS NVMe volumes rather than throttled network-attached disks—giving PostgreSQL 17, MySQL 8.4, and Redis 7 consistent sub-millisecond query latency.

DB-02 < 1s clone

Instant Copy-on-Write Branching

Fork a 50 GB database into an isolated branch in under a second without duplicating storage blocks. Give every pull request or AI coding agent its own realistic test database.

DB-03 WAL Archiving

Continuous Point-in-Time Recovery

Automated WAL streaming and scheduled snapshots let you restore your PostgreSQL or MySQL instance to any second inside your retention window.

DB-04 Auto Failover

Optional High-Availability Replicas

Add synchronous or asynchronous standby replicas for production workloads requiring automated leader election and read-scaling without manual cluster surgery.

DB-05 $0.00/hr idle

Sleepable Dev & Preview Branches

Production databases stay pinned 24/7 while dev, staging, and PR branches automatically pause to $0/hr compute when tests finish—paying only for delta storage blocks.

DB-06 Zero public exposure

Native Private Network & TLS

Connect over private project DNS from your uncage.sh apps, sandboxes, and CI runners with zero egress fees, or toggle encrypted external access when needed.

REAL-WORLD WORKFLOWS

Stop testing migrations on empty mocks or shared staging DBs.

AI CODING AGENTS

Isolated database branch per agent trajectory

The Pain

Letting an AI agent run destructive schema migrations or seed scripts against a shared dev database corrupts state for the entire team.

With uncage.sh

Call fork() on your staging Postgres database inside the agent loop, pass the branch URI to the sandbox, and discard or inspect after verification.

→ 100% safe autonomous database testing with sub-second setup time.
CI & MIGRATIONS

Verify Alembic, Prisma, or Flyway migrations against real volume

The Pain

Migrations that pass on an empty CI Postgres container lock tables or fail on edge-case production data shapes.

With uncage.sh

Branch a scrubbed production snapshot in CI, run your migration + integration suite on dedicated NVMe, and tear down automatically.

→ Zero migration rollbacks in production; no idle staging DB tax.
PRODUCTION OLTP

Predictable high-IOPS Postgres + Redis without cloud markup

The Pain

Managed cloud databases charge steep IOPS add-ons, per-branch fees, and high cross-AZ network transfer costs.

With uncage.sh

Run PostgreSQL 17 and Redis 7 co-located with your application servers in us-east, eu-central, us-west, or ap-south at transparent active rates.

→ Up to 60% lower monthly database spend with faster disk throughput.
TRANSPARENT DATABASE PRICING

No hidden IOPS surcharges or per-branch seat fees.

Databases use the exact same compute + disk pricing: $0.036/vCPU-hr, $0.0054/GB-hr, and $0.10/GB-mo for dedicated NVMe storage. Branch clones only bill delta blocks written.

Open Interactive Pricing Calculator →
Dev / PR Branch Postgres 17
1 vCPU · 1 GB · 1 GB delta
2 hrs/day active, auto-sleeps otherwise
Estimated $2.58 / mo
Production Redis 7 Cache
1 vCPU · 4 GB · 10 GB NVMe
24/7 active (730 hrs/mo)
Estimated $43.05 / mo
Production PostgreSQL 17
4 vCPU · 8 GB · 40 GB NVMe
24/7 active (730 hrs/mo)
Estimated $140.66 / mo
Database Compute Spec Dedicated NVMe Active Rate Sleeping Dev Branch Provision
1 vCPU · 1 GB 10 GB ($0.10/GB-mo) $0.0414/hr $0.0000/hr Launch →
2 vCPU · 4 GB 20 GB ($0.10/GB-mo) $0.0936/hr $0.0000/hr Launch →
4 vCPU · 8 GB 40 GB ($0.10/GB-mo) $0.1872/hr $0.0000/hr Launch →
8 vCPU · 16 GB 80 GB ($0.10/GB-mo) $0.3744/hr $0.0000/hr Launch →
MANAGED DATABASES FAQ

Technical details on NVMe databases & copy-on-write branching.

01 How fast is copy-on-write database branching for large databases?
Branching creates an instant snapshot clone at the storage layer without copying existing data pages. Whether your database is 500 MB or 100 GB, a writable branch comes online in under a second and only consumes additional disk for blocks modified by the branch.
02 Which database versions and extensions are supported?
Managed templates ship with PostgreSQL 17 (including pgvector, pg_stat_statements, and PostGIS support), MySQL 8.4 LTS, and Redis 7. You can also specify custom minor-version OCI images via the CLI or API.
03 Can production databases stay always-on without ever sleeping?
Yes. Production databases default to always-on pinned execution with zero idle sleep, dedicated NVMe storage, and continuous WAL backup streaming. Auto-sleep is opt-in and ideal for dev, QA, and PR preview branches.
04 How does point-in-time recovery (PITR) work?
Write-ahead logs (PostgreSQL WAL / MySQL binlogs) are continuously archived alongside periodic base snapshots. You can restore a new database instance to any timestamp within your retention window without overwriting your live primary.
DEDICATED NVME DATABASES

Launch PostgreSQL 17, MySQL 8.4, or Redis 7 in seconds.

Get local NVMe performance, continuous backups, and instant branching in the same workspace as your apps and agents.

✓ $10 starter credit on signup
✓ Strictly $0.00/hr compute when sleeping
✓ Sub-second boot across 4 global regions