Skip to content
Foundation · On-prem agent

Runs where the database lives.

A small service you install on Windows or Linux, holding the real connection string locally and opening an outbound TLS channel to the cloud. It runs the documentation scans, the queries somebody types, and the exports that go out on a schedule. Nothing ever comes back the other way.

Where it runs

Windows, Linux, Docker.

A bundled installer registers it as a Windows service, a systemd unit covers Linux, and there is a Docker image with a documented compose file.

How it talks to the cloud

Outbound TLS only.

It holds a persistent connection over HTTPS and never listens on anything, so the only firewall rule involved is the outbound 443 you already have.

How it holds credentials

Sealed envelope.

The browser seals the connection string with the agent public key before it leaves, the cloud only ever stores that sealed envelope, and the agent is the one thing able to open it.

What the agent does

Three jobs, all read-only.

Job 01

Schema scans

Read information schema and provider catalogs to feed the documentation and the change history.

Job 02

Ad hoc queries

Execute saved or freeform SELECT statements you authored, under read-only role and timeout.

Job 03

Scheduled exports

Run a saved query on schedule, format as CSV or Excel, push it to the cloud for delivery to Slack or email.

Capacity

One agent, many connections.

A team that runs three databases does not need three agents. Concurrency caps protect the agent host and the customer database from spikes during scheduled run windows.

Per agent

32 concurrent runs

Default cap on a 4 vCPU host. Tunable.

Per connection key

8 concurrent runs

Avoid hammering one database with parallel diagnostics.

Schedules

Hourly cadence on Pro

Per-connection schedule. Run-now button when you do not want to wait.

FAQ

On-prem agent for private databases

What ports does the agent need? +

Outbound 443 to the control plane. Nothing inbound. The agent opens a long-lived SignalR connection and keeps it warm with heartbeats.

Can the agent run on the same host as the database? +

Yes. Many teams run it as a sidecar on the database host or on a small jump host with network access to the database.

How is the agent updated? +

Replace the binary, restart the service. The agent reports its version on connect; the cloud blocks versions below MinAgentVersion.

Can I run multiple agents for redundancy? +

Yes. Multiple agents can be enrolled. The cloud picks one for each scheduled run via a lease. Two agents racing to scan the same database will not duplicate work.

Install it

Drop the agent on a host. Start documenting in minutes.

Windows, Linux or Docker, outbound on 443, and the first scan finishes in under ten minutes.