Describe what to test in plain language — then keep watching the same targets from your nodes with latency and failure alerts.
No credit card · icmp · tcp · http · traceroute · dns
capabilities
Diagnose in chat, then turn the same vantage points into continuous checks.
Ask in English. The agent picks icmp, tcp, http, dns, or traceroute.
Self-host probes with systemd. Traffic leaves from networks you trust.
Schedule jobs on a few-minute cadence. Charts for latency & TTFB; alerts on failure.
continuous monitoring · 主动监控
Turn a diagnosis into a job. Track HTTP latency & TTFB, or ICMP RTT & loss, with clear thresholds.
03:14–03:18 HKT — brief latency spike (p95 ≈ 480ms) then recovery. No sustained failure; alert threshold not crossed.
how it works
Aligned with ConnSentinel's diagnose → monitor flow.
Reuse the host and region you just diagnosed — or choose another probe in your fleet.
HTTP checks every 30s–10m. ICMP as fast as 15s. Tune for cost vs. signal.
Notify when checks fail, or when p95 / loss crosses the threshold you set.
security · self-host
Your host dials out. Nothing listens for the public internet. Same vantage points power chat diagnosis and continuous monitoring.
No public listener, no open firewall holes. The probe only makes outbound HTTPS to the agent registry.
Enrollment uses a short-lived token tied to your account. After check-in, the node appears in your fleet — nothing else.
Runs as a systemd service with CAP_NET_RAW for ICMP where needed, and NoNewPrivileges — not a full root shell.
alerts
Failure alerts and latency thresholds, without drowning you in noise.
Open the app to diagnose now — monitoring jobs unlock when you're signed in.