Company News
Hosted Postgres for your AI agents
AI agents can now create a hosted PostgreSQL database through DBGorilla's MCP server, with monitoring running from the moment it starts.
Cornelius·
An AI agent can now create a PostgreSQL database through DBGorilla’s MCP server and start using it with monitoring already running. DBGorilla hosts the database for your workspace and monitors it from the moment it starts. You have nothing to install.
Creating and managing databases
The MCP server lets agents create, list, inspect and delete databases, and issue logins. Creating or deleting a database and issuing a login require write permission on the workspace. To delete a database, the agent must repeat its exact name.
People can perform the same actions from the Databases page in the app or through the API.
Database configuration
Choose PostgreSQL 14 to 18. If you do not choose a version, DBGorilla uses the newest.
| Size | CPU | Memory | Storage |
|---|---|---|---|
| Small | 0.5 CPU | 1 GiB | 5 GiB |
| Medium | 1 CPU | 2 GiB | 20 GiB |
| Large | 2 CPU | 4 GiB | 50 GiB |
Each database runs as a CloudNativePG cluster on Kubernetes. It runs one to three instances: a primary, with standby replicas that keep the database available during maintenance.
DBGorilla installs and preloads pg_stat_statements. Databases run around the clock, without hibernation or expiration.
Logins and network access
Every login is short-lived and connects through DBGorilla’s access proxy. The agent receives a connection URI that requires the client to verify the proxy’s TLS certificate. DBGorilla shows the password once and does not store it. When the login expires, the agent requests a new one.
Each workspace has its own Kubernetes namespace for its databases. Only a database’s own instances, the CloudNativePG operator, DBGorilla’s collectors and the access proxy can reach that database. Nothing in another workspace can reach it.
To create a database, open Getting started in DBGorilla and choose Create your database.