Guide for AI coding agents

PostgreSQL for AI agents.

Give an agent a real database it can use for migrations, integration tests and prototypes, then remove it automatically when the task is done.

The problem

Agents need a database they can actually use

AI coding agents are good at changing schemas, writing queries and running test suites. They still need an isolated PostgreSQL environment to verify that work. A shared development database creates collisions, stale data and cleanup work.

Aiadio creates a real temporary PostgreSQL database through an API. The agent receives a standard connection URL, uses normal Postgres tools, and the database is removed when its time-to-live expires.

Useful default:

Give each task its own database and choose the shortest TTL that covers the run. This keeps agent work isolated and limits forgotten environments.

Agent workflow

One database per task

  1. Create a temporary database before the agent starts its database work.
  2. Pass the returned PostgreSQL URL to the agent as a protected secret.
  3. Let the agent run migrations, seed data and integration tests with standard tools.
  4. Read the result and keep or discard the application changes independently.
  5. Let TTL cleanup remove the database and its role automatically.

This works for coding agents, CI jobs, preview environments and short-lived prototypes. It does not replace a durable production database.

Quickstart

Create temporary Postgres by API

Use an Aiadio API key in the X-API-Key header. Keep the key and the returned connection URL server-side or in your agent runtime's secret store.

Create a database
curl -X POST https://api.aiadio.com/v1/databases \
  -H "X-API-Key: dbn_live_your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "agent-task-42",
    "expires_in": "3h",
    "allowed_cidr": "203.0.113.10/32"
  }'
Response shape
{
  "id": "db_8f4c2",
  "status": "ready",
  "postgres_url": "postgresql://...",
  "expires_at": "2026-09-16T15:00:00.000Z"
}

Use postgres_url with PostgreSQL clients, ORMs, migration tools and test runners. The API is documented in full on the Aiadio API docs.

Operational boundaries

Keep agent access narrow

  • Use a separate Aiadio API key for the workload and keep it out of prompts and source control.
  • Restrict network access with the smallest practical allowed_cidr.
  • Never expose a returned PostgreSQL URL in browser JavaScript or public logs.
  • Use short TTL values for experiments and test runs; use a durable database for data that must persist.
FAQ

Postgres for AI agents: common questions

Is this a mock database?

No. Aiadio provides a real PostgreSQL database that works with standard PostgreSQL clients and tooling.

Can an agent run migrations and tests?

Yes. Give the agent the returned PostgreSQL URL through a protected runtime secret and point your migration or test command at it.

How long does the database live?

The documented TTL range is 1 hour to 30 days. When the TTL expires, Aiadio removes the temporary database and role automatically.

Where can I see the exact request fields?

See the API documentation for authentication, create, status and error responses.

Build the agent workflow around an isolated, disposable Postgres database.

Get started