Service status

Where to check

• Status page: status.inboxy.net (lands at v0.3a — until then, see the GitHub repo's CI badge).
• Incidents in flight: posted to the status page and, for SEV-1, the @inboxystatus social handles.
• Historical incidents: the status page archives every published incident going back to launch.

Components

Component	What it does	What "down" looks like
Email in (email-in worker)	Receives mail at *@inboxy.net	New mail not arriving; senders see a hard bounce
Classifier	OTP / receipt / marketing tagging + summarisation	New mail arrives but lands in "Awaiting classification"; existing mail still searchable
Sender	Outbound digests + forwards	Digests not sent at their scheduled time
API / MCP	Programmatic access	Agents and the extension get 5xx
Web	inboxy.net + the account UI	This page wouldn't load
Link	Click-redirects on l.inboxy.net	Clicking a rewritten link 5xxs (forward and digest mail include rewritten URLs)
Digest	Daily digest scheduling	Same as sender if upstream — but digest can be down while sender is healthy if the scheduler itself is wedged

Degraded modes

We try to degrade gracefully:

• Classifier down → mail is delivered raw, tagged awaiting classification. You can still search and read.
• Sender down → digests queue up. When sender recovers it drains the backlog (idempotent on each digest run, so you don't get duplicates).
• API down → the web UI may still work (different worker, different bindings). The MCP server is also separate.

SEVs

We use four severity levels internally:

SEV	Trigger	Response
SEV-1	User-visible outage of a core component for > 5 min	Pager, status page within 5 min, post-incident review
SEV-2	Partial degradation or single-tenant impact	Status page within 30 min
SEV-3	Internal-only or near-miss	Tracked; not surfaced unless asked
SEV-4	Telemetry/observability gap	Backlog only

Notifying us

If you think we're down and the status page doesn't reflect it: support@inboxy.net. Mention your personal_email so we can correlate; we don't need the rest of your account details to investigate a service outage.