Security

Audit log

The append-only record of every sign-in, change, denial and operation in VaultPanel, what each entry holds and how to filter it.

Last updated

The audit log records who did what, to what, from where, and whether it worked. It covers sign-ins, every change made through the panel, the API, the CLI, plugins and the AI assistant, permission denials, and events the panel raises itself, such as alerts firing and servers going offline. Entries cannot be edited, and are removed only after the retention period you choose.

Read the audit log

Open Security > Audit Log. It needs audit.read (owners and admins by default). Entries are listed newest first; click one to see all its details.

Filter the list by:

  • Action: entries whose action starts with the text you type, for example server. for everything about servers or auth.login for sign-ins.
  • Result: Success, Failure or Denied.

The API also filters by actor, target type and ID, and a time range.

What an entry holds

FieldWhat it is
WhenThe time of the action
WhoThe actor: a signed-in user (by email), an agent, a plugin, a billing API key, someone not signed in yet, or the system itself
ActionWhat was done, named area.verb, such as website.create
TargetWhat it was done to: a type, an ID and a label
Resultsuccess, failure or denied
IPThe address the request came from
User agentThe browser or client
Request IDLinks the entry to the panel's server log and to the API response
DetailsMetadata about the change, such as the before and after of a setting

A user acting through an API token is recorded as that user. Actions the AI assistant runs are recorded as the person who asked, labelled email (assistant), and plugin actions as plugin:<id>.

What is recorded

AreaExamples
Sign-inauth.login (with the reason for failures), auth.logout, auth.password_change, auth.session_revoke, auth.setup_owner
Accessauthz.denied for every refused permission, user.invite, user.update, api_token.create, api_token.revoke, security.two_factor_policy, security.panel_access
Serversserver.create, server.delete, server.agent_update, server.offline, server.online, server.terminal.open, server.terminal.close
Jobsjob.create, job.cancel, and the server security changes they carry, such as security.firewall_rule_add and security.firewall_enable
Monitoringalert.fire, alert.resolve, alert.acknowledge, alert_rule.create
Hostingwebsite.*, application.*, database.*, backup.*, cron_job.*, dns_zone.*, mailbox.*, container.*
Secrets viewedapplication.env_view, database_user.password_view, container.env_view, website.env_view
Businesscustomer.*, plan.*, api_key.create, api_key.revoke
Extensionsplugin.*, webhook.*, assistant.action_approve

Secrets are not written to the log: passwords, environment variable values and file contents stay out of entries, and are replaced with a redacted marker in a job's stored parameters when the job ends.

Protection

On MariaDB and MySQL, database triggers protect the audit log table:

  • No entry can be changed, with one exception: when a person is erased, the name or email shown becomes "Erased person" and the IP address and browser are removed. What was done, by which account, when and with what result never changes.
  • No entry can be deleted until it is older than the retention period, and never before it is 90 days old, whatever the panel asks.

No API can change entries. TRUNCATE and DROP bypass triggers, so keep direct access to the panel's database restricted.

If writing an entry ever fails, the failure is logged loudly, but it never hides the outcome of the action being audited.

Retention

By default entries are kept forever. Owners and admins can choose a period at the top of the Audit log page: 90 days, 6 months, or 1, 2, 5 or 10 years. The background worker then removes older entries, a batch at a time, and the change of period is itself recorded. Pick a period that fits your security and legal needs and state it in your privacy notice; see Personal data and GDPR.

Events from the audit log

Some entries also become events you can receive elsewhere: an alert firing or resolving, and a server going offline or coming back online, can be sent to webhooks and by panel email.