You'd think this needs a status page review and three Datadog dashboards — Tycoon Agent has the weekly health summary plus the one incident that actually cost you money.
The short answer
Tycoon Agent answers "how is my system uptime this week" by pulling availability metrics from Vercel, Cloud Run, Datadog synthetic checks, and Sentry error rates, then computing real uptime per critical service over the trailing 7 days. She doesn't just report a single nine-count — she breaks it down by surface (marketing site, checkout, API, dashboard) and weights each incident by what it touched. A 12-minute Cloud Run blip during European business hours that hit your signup endpoint matters more than a 2am marketing-site hiccup. The output is a Lark message: per-service uptime percentage, total downtime in minutes, the worst incident with root cause from Sentry, and an estimate of users affected pulled from PostHog session counts during the window. You stop chasing alert noise and focus on the one outage that actually mattered.
How Tycoon Agent actually does it
- Pull availability per service
Vercel deployment health, Cloud Run request success rate, Datadog synthetic uptime checks, Sentry error rate spikes — last 7 days.
VercelDatadogSentry
- Weight by surface criticality
Checkout > API > dashboard > marketing. Same downtime gets a different severity depending on what broke.
- Cross-reference with PostHog traffic
Match incident windows to PostHog session counts to estimate users actually affected, not theoretical worst case.
PostHog
- Identify the worst incident
Pull Sentry stack trace and Vercel deploy log around the incident window, surface root cause and which deploy introduced it.
- Send the weekly digest
Lark message: per-service uptime, total downtime, worst incident with root cause, users affected, fix status.
Lark
What it looks like in chat
YouHow is system uptime this week?
Tycoon AgentTrailing 7 days: - **Marketing site (Vercel):** 99.98% — 2 min cold-start blip Tuesday 3am. No users affected. - **API (Cloud Run):** 99.81% — 16 min total downtime. Two incidents. - **Dashboard:** 99.94% — 5 min Sentry error spike Thursday 9pm. - **Checkout:** 100%. **Worst incident:** Wednesday 14:32-14:48 UTC. Cloud Run API returned 503 on `/v1/run` for ~16 min. Root cause: deploy `a3f9c2` removed an env var without a fallback. PostHog shows 84 active sessions in that window, ~23 failed requests. Auto-rolled-back at 14:48, fix deployed 15:02. **Action:** Add a startup check for required env vars to your Dockerfile (~30 min). I can file the task and assign to your developer agent. Otherwise this exact failure repeats next time someone removes a config key.
What you get back
Lark digest every Monday with per-service uptime, total downtime in minutes, the single worst incident with root cause and users affected, plus a concrete fix recommendation.
Cadence: Weekly Monday + immediate alert on any 5+ min critical-service outage.