Deliverability Agent

Watches send-level health metrics on every cron tick. Flags reputation risks before they become inbox-folder problems and auto-pauses active campaigns when ESP red-lines are breached.

What it watches

For each workspace, computes over the last 24h of sends:

• Bounce rate — bounced / total sends
• Complaint rate — complained / total sends

A workspace must have at least minSendVolume sends in the window before any metric is eligible to signal (default 100). Below that, rates are too noisy.

Auto-pause

When either rate exceeds its threshold, all active campaigns in the workspace are flipped to paused and a critical insight is filed. Defaults match ESP red-lines:

• bounceRatePauseThreshold = 0.03 (3%)
• complaintRatePauseThreshold = 0.001 (0.1%)

All three thresholds are per-workspace under /dashboard/settings/workspace → Agent thresholds. Tighten for regulated workspaces (HIPAA: 1% bounce typical); loosen for known-clean transactional senders.

Auto-suppression

Independent of the sweep, hard bounces auto-suppress on the first event via the Resend webhook handler. Soft bounces are off by default but can be enabled per workspace with a consecutive-failures threshold + lookback window (delivery.softBounceAutoSuppress in the workspace settings). A single successful delivery within the window resets the streak.

Recovery

When the agent auto-pauses a workspace, the insight summary lists the affected campaigns + the breaching rates. Resuming requires a human action in the dashboard — by design.

Schedule

Every 10 minutes via /api/cron/deliverability. Manual trigger from /studio/inspector.