Always-on database
Dedicated PostgreSQL for production
Always-on databases run on dedicated resources around the clock, for the applications your agents ship and the services that must respond at any moment.
What a production database needs
Standard PostgreSQL, operated for you.
High availability
Replicas on separate machines take over if the primary fails, and the connection address follows the new primary.
Connection pooling
A built-in PgBouncer pool runs in transaction mode by default, with session mode available.
Backups and point-in-time recovery
Back up every 6, 12, or 24 hours, keep backups for 1–30 days, and restore any point in that window to a new database.
Storage autoscaling
Storage grows automatically when usage reaches 80%, up to a maximum you set.
Parameters and extensions
Tune PostgreSQL parameters and turn on extensions such as pg_stat_statements.
Online cluster changes
Add or remove replicas without a restart. A plan change restarts the database once.
Plans
Every node is billed at the plan price, so a highly available database is billed for its primary and each replica.
| Plan | vCPU | Memory | Per node / month |
|---|---|---|---|
| S-1 | 1/16 | 512 MB | $5 |
| S-2 | 1/8 | 1 GB | $9 |
| S-4 | 1/4 | 2 GB | $16 |
| S-8 | 1/2 | 4 GB | $29 |
| S-16 | 1 | 8 GB | $48 |
| S-32 | 2 | 16 GB | $95 |
Each node includes 10 GB of storage. Additional storage is $0.375 per GB-month on each node.
Specifications
- Topology
- Single node, or 1 primary with 2–10 replicas
- Storage
- 10–1,000 GB
- PostgreSQL
- Versions 17 and 18
- Connections
- Pooled on port 6432; direct on port 5432
- Grantable roles
- owner, pg_read_all_data, pg_write_all_data, pg_read_all_stats, pg_monitor, pg_signal_backend, pg_maintain
- Regions
- cn-hangzhou (Hangzhou), cn-beijing (Beijing), cn-shanghai (Shanghai), cn-shenzhen (Shenzhen), ap-east-1 (Hong Kong), ap-southeast-1 (Singapore)
Common uses
Always-on databases suit workloads that must respond immediately at any time.
Production applications
Serve the applications your agents build and deploy with consistent performance.
Agents that never stop
Back agents that run continuously and query their data at every step.
Many short connections
Serve large numbers of short-lived connections from many application instances through the pool.
Need a database only some of the time?
On-demand databases scale with load and pause when no one is connected.
Learn more
Create the first database for your agents
Sign up and create an API token to begin.