Ask for what you need. Tern does the wiring.

Describe the change in one sentence. Your coding agent finds the events, writes the tracking code and puts the chart where you asked, as plain files in your repo.

web — agent

localhost:3000/alerts

Alerts

2 rules
Signup drop−30% in 1 hour

Nothing watches checkout yet

Slack
Not connected
  1. You: Ping #payments in Slack when checkout errors spike
  2. Ran Bash: npx tern events list --match checkout. checkout_started checkout_failed checkout_completed
  3. Ran Bash: npx tern alert add checkout_failed --above 2% --window 10m --slack "#payments". + tern/alerts/checkout-failed.yml 2% over 10 min. ✓ Slack connected · #payments
  4. Ran Bash: npx tern alert test checkout-failed. ✓ Test alert sent to #payments
  5. Agent: Done. If more than 2% of checkouts fail within ten minutes, #payments hears about it.

Every change is a file you read in review and undo with git.