---
title: Connect
description: "Find the connection string, host and port of a database service, in JDBC, URI or psql format, and the other endpoints of an HA service."
---

On the **Overview** tab, select **Connection info** in the **Topology**
header. The panel shows the connection strings. Select **JDBC** (default),
**URI**, or **psql** to change the format.

![The Connection info panel with the format selector. An outline marks the recommended endpoint.](/docs/screenshots/v2026.11.6-beta/connect/connection-info.light.webp)

The user is `postgres`. Below the admin role, the connection strings do
not show the password.

## Endpoints

| Endpoint | Standard | High-Availability |
| --- | --- | --- |
| **Recommended endpoint** (use it for applications) | PgBouncer `:6432` | HAProxy `:5000`, pooled through PgBouncer on the current primary |
| Primary, not pooled | PostgreSQL `:5432` | HAProxy `:6000` |
| Primary, direct (**Other endpoints**) | PostgreSQL `:5432`, PgBouncer `:6432` | PostgreSQL `:5432`, PgBouncer `:6432` on the current primary |

On a High-Availability service:

- With a VIP, the `:5000` and `:6000` endpoints are on the VIP only.
- Without a VIP, every node has `:5000` and `:6000`, and the connection
  string names all nodes.
- The connection string adds `target_session_attrs=read-write`, so the
  client connects to the primary.
- **Connection info** does not show an endpoint of a replica. A replica is
  read-only, and it can be behind the primary.

In a demo, the recommended endpoint goes through the computer that runs the demo. See
[Demo mode](/docs/v2026.11.6-beta/demo-mode#connect-to-a-database-service).

See the [connection path](/docs/v2026.11.6-beta/reference-architectures#connection-path) for
the full port map.

:::note
A direct endpoint names one server. After a failover, the primary is on a
different server, and **Connection info** shows the direct endpoints of the
new primary. Use the recommended endpoint for applications.
:::
