Prerequisites
- An InboxMate account — sign up for free
- A mailbox you can connect — Outlook / Microsoft 365 is the fastest path; email forwarding works with any provider
Connect your Postfach
1
Log in and open Mein Profil
Go to app.psquared.dev and sign in, then open Einstellungen → Mein Profil → Meine Postfächer.
2
Click Postfach hinzufügen
Pick Outlook / Microsoft 365 for one-click OAuth — sign in with your Microsoft account and approve the requested permissions. Works with both M365 work accounts and personal Outlook.com / Hotmail accounts.Prefer a different provider? E-Mail-Weiterleitung works with Gmail, IONOS, FastMail, or anything that supports a forwarding rule; SMTP/IMAP is available by contacting sales. See Email integration for the full comparison.
3
Wait for the first sync
New mail starts syncing within about 2 minutes. You can trigger an immediate pull from the connection’s settings popover if you don’t want to wait.
On a managed Microsoft 365 tenant, your admin may need to approve InboxMate once before anyone can connect — you’ll see an “Administratorgenehmigung erforderlich” screen if so. See Microsoft 365 admin consent for both ways past it.
Confirm your categories
Right after your mailbox connects, InboxMate offers to set up categorization so the very first email that arrives is already sorted: There is nothing to choose here. The step starts scanning the moment it opens — it waits for the first mailbox sync to finish, reads up to 100 of your most recent emails, and proposes a small set of categories that match your real traffic. Each proposal is an editable card:- Rename the category or rewrite its one-line description
- Check the Beispielsender chips — the actual senders the AI saw for that category
- Delete a category you don’t want, or add your own with Kategorie hinzufügen
If your mailbox is brand-new and empty, you get an “empty inbox” notice and a E-Mails erneut laden & analysieren button. If the scan takes too long, InboxMate falls back to a proven starter set (Wichtig · Newsletter · Spam) that you can edit the same way — you’re never blocked. If the request fails outright you get Analyse hat nicht geklappt, with Erneut versuchen and Mit Standard-Kategorien starten.
Full autonomy is earned, not preset. To start, InboxMate only suggests actions — you approve them. The AI learns from your approvals and corrections; once a rule has matured you can switch it to autopilot in the KI-Reife tab.This is the part people expect to be a dial and it isn’t. Every action created during onboarding is suggest-only, without exception — including archiving and spam. Nothing your new categorizer does will fire on its own until you graduate it yourself. That’s why each action chip on the category cards reads vorschlagen. Click Übernehmen und weiter. InboxMate saves one categorizer scoped to the mailbox you just connected.
Watch it work
The last onboarding step runs the first live triage in front of you: it pulls your newest mail, categorizes the most recent conversations from the last 14 days (up to 15 of them), and reports how many emails it triaged. Then Zu deinen Entscheidungen takes you into the approval queue. If your mailbox is quiet and there’s nothing to sort yet, you’ll see Noch keine neuen E-Mails — your categories are already active, so mail that arrives later gets sorted automatically. Open your inbox afterwards and each conversation shows a category badge, and any category configured to draft a reply already has one waiting, written from your knowledge base, the contact’s history, and your signature.Review and approve in Entscheidungen
Open Entscheidungen in the sidebar. Every suggestion InboxMate made — a draft reply, an archive, a category assignment — sits here until a person acts on it:- Ja — approve it as-is
- Anders — adjust it before it goes through
- Ignorieren — dismiss it
Next steps
Entscheidungen
How the approval queue works, and what each action type looks like.
Learning
How your edits and approvals actually change what InboxMate does next.
AI maturity
Track how much trust InboxMate has earned, category by category.
AI models
Which models power drafting and categorization, and how to configure them.