Audit log — who did what, when

The audit log is the answer to "who did what, when?" — every admin-side action that mutates state is recorded with the actor, the target, the timestamp, and a JSON metadata blob. Use it to investigate a misconfiguration after the fact, to satisfy a regulator who needs proof of access controls, or to track what an automation did overnight.

What's recorded

A non-exhaustive list of action categories that land in the log:

  • Tool — created, edited, deleted, secret rotated, tested.
  • Document — uploaded, deleted, reprocessed.
  • Conversation — notes updated, snippet created from a turn, handoff resolved.
  • Snippet / knowledge — created, edited, deleted; merge-proposal decisions.
  • Team — invites sent, revoked, members removed, permissions changed.
  • Settings — SSO config saved, widget config changes, billing actions, data-export requests.
  • Compliance — DPA requested, sent, canceled; subprocessor list acknowledged.

Read-only browsing (looking at the leads list, opening a settings page) is not logged, and neither are sign-ins — the log covers state-changing admin actions only.

Filtering the log

The filter bar lets you narrow by:

  • Action — pick from the Action dropdown (tool.create, document.upload, etc.). The list is fixed; you can't search for ad-hoc patterns.
  • Date rangeFrom / To date fields for a custom range.
  • Free-text — searches the actor email, the entity id, and the action name. It does not search the JSON metadata.

The page shows 50 entries at a time; pagination at the bottom.

What's in each row

  • When — exact timestamp, in your local time.
  • Actor — the email of the person (or service) who did the action. SSO-provisioned members appear with their IdP-asserted email. System actions (cron jobs, automated cleanups) appear as system.
  • Actionnoun.verb style; tool.create, team.remove_member, tenant_dpa.send.
  • Entity — what was acted on. Usually a name or id; click into the row for the raw JSON.
  • Metadata — additional context (before/after values, which field changed). Some historical entries have empty metadata; we've been gradually filling it in.

Export to CSV

The Export CSV button dumps the currently-filtered view to a CSV. Useful for:

  • Sharing with your auditor or security team.
  • Importing into a spreadsheet for ad-hoc analysis.
  • Long-term retention beyond what we keep on the page.

The button is always shown, but the export itself has to be enabled on your workspace. If clicking it gives you a "not enabled for this workspace" error, email us to enable.

Retention

Log retention is a fixed 365 days for every workspace. The page surfaces a retention note — anything older than that window has been purged. If you need longer retention, export periodically.

Common questions

  • Who can see the audit log? Currently any member can read it. Sensitive metadata (passwords, secrets) is never logged in the first place.
  • Where do I see who signed in? You don't — sign-ins (password or SSO) aren't audited. The log records state-changing admin actions only.
  • Can I export beyond the displayed window? Only what's in the current filter. Filter by your desired date range first, then export.
  • Why doesn't action X appear? Either it's a read-only browse (not logged by design) or it's a recent feature where metadata is still being filled in. If you think a write was missed, file a bug.
  • How long is the log kept? 365 days, for every workspace; the retention note on the page says the same.
  • What's a system actor? A scheduled job or an automated cleanup did the action — no human pressed a button. E.g. retention purge, webhook retry.

See also team for permissions and compliance for DPA actions in the log.

Where to find us

Stuck? Email wilow@werila.com. If you need the audit-log CSV export turned on, mention that.