Pre-launch. No uptime to report yet.

Nine systems, listed below. Once LedgerInbox is serving real books, each one gets a 90-day availability history here, and every incident gets a post-mortem with a timestamped update log — including the ones that are our fault.

• Categorization inbox
  Web app · keyboard inbox · ⌘K palette
  No data yet
• Report ingestion
  Per-location email intake · POS and payroll parsing
  No data yet
• Reconciliation engine
  Cash over/short · tip rollforward · deposit matching
  No data yet
• QBO write-back
  OAuth · journal entries · closed-period guard
  No data yet
• AI categorization
  OpenAI gpt-4o-mini · transaction line only, never balances
  No data yet
• Receipt OCR
  In-browser read · auto-match queue
  No data yet
• Authentication
  Supabase magic links · session cookies
  No data yet
• Database
  Postgres · per-firm row-level security
  No data yet
• Plaid bank sync
  Optional · item refresh · transaction webhooks
  No data yet

1. An incident is posted here before it is explained to anyone privately. If your client noticed it, you should not be hearing it from them first.
2. Degraded counts as an incident. Slow is a failure mode on a product whose entire claim is speed, so it gets the same treatment as down.
3. Post-mortems name the cause, including when it is us. No “an upstream provider experienced an issue” when the upstream provider was our own deploy.
4. No uptime figure appears here until it is measured. This page carried invented numbers once. That is why it now carries none.