Know it went down before your customer does.
Checks from our infrastructure in Brazil, with intervals starting at 1 minute.
Eight observation modes — from HTTP and keyword to a login flow in a real browser. Alert in seconds, on the channel you choose.
No card · 1 free monitor forever · 2-minute setup
example.com is down — HTTP 502, confirmed in 2 checks.
Observation modes
Eight modes, one for each case.
Seven in pullmode — we fetch the target. One in pushmode, for an internal job with no exposed HTTP. And synthetic, which opens a real browser and runs the user flow end to end.
HTTP(S)
Checks the expected status (200, 301) within the timeout. On HTTPS, watches the certificate and warns 30 days before expiry.
What for
Home page, REST API, admin panel, payment webhook.
Keyword
Extends HTTP by looking for — or ruling out — a keyword in the response body.
What for
Error page returning 200, unplanned maintenance banner, defacement.
Ping (ICMP)
Sends an echo packet to the host and measures the response time.
What for
Server with no exposed HTTP, gateway, network equipment.
Port (TCP)
Opens a socket on a specific port, for a service that does not speak HTTP.
What for
SSH (22), SMTP (25), MySQL (3306), Redis (6379).
Heartbeat (push)
Inverts the flow: your cron or worker pings a unique URL. If it does not arrive within the interval, with 2x slack — it goes DOWN.
What for
Nightly backup, data sync, queue worker, internal job with no exposed HTTP.
DNS (resolves A)
Resolves the host A record. Up if at least one address comes back — catches a lost zone and slow propagation.
What for
Confirm resolution after switching NS, validate a client external zone.
Vercel (error logs)
Queries the official Vercel API on every check. Goes DOWN when the production deploy fails or runtime errors pass your threshold.
What for
Next.js app on Vercel: catch a broken deploy or an exception spike before the customer complains.
Synthetic (login/flow)
A real browser runs the multi-step flow: opens, fills, clicks, waits, validates text or URL — with encrypted login.
What for
Panel login, checkout, search, authenticated area. Per-step SLO and a screenshot of the failure. Pro+.
Notifications
Eight channels — pick where you get it.
Each monitor can fire to several channels at once. Set it up once, reuse it everywhere.
The default for everyone. Clean layout, direct link to the incident, several recipients per channel.
Webhook
JSON with the full payload and custom headers. Connect n8n, Zapier or your own system.
Telegram
Official bot, set up in 30s by pasting the chat ID. Great for a phone alert without an extra app.
Slack
Rich message in the team channel, colored by severity. Official Slack webhook.
Discord
Colored embed in the server. Ideal for a team that already lives on Discord.
PagerDuty
Creates the incident when the monitor goes down and resolves it when it comes back. For serious on-call.
Opsgenie
Same model as PagerDuty: alert on the outage, resolve on recovery. For Atlassian teams.
Jira
Opens an issue in your project when it goes down, comments and transitions to Done when it recovers. Token or OAuth.
Public status page
Transparency without friction.
Every incident tells a story. Your status page tells the whole story — what went down, when it came back and the uptime of the last 30, 60 or 90 days.
Maintenance windows
A planned deploy is not an alert.
Taking the API down at 3 a.m. on Saturday to run a migration? Create a window and the monitor understands.
- Does not fire an alert during the window
- Does not count the time against your uptime%
- Shows “scheduled maintenance” on the status page
Auto-pause
Client gone? It pauses itself.
The client cancelled and nobody told you. The domain has been down for days and your panel becomes visual noise — unless the monitor pauses itself.
- Limit in hours, adjustable per monitor (empty = off)
- The audit trail records the pause
- Manual reactivation resets the cycle
Latency
p50, p95, p99 — the truth about the slow ones.
The average lies. At p95 you see how long the twentieth-worst user waited; p99 is the worst in a hundred. We compute it over every check in the interval and compare week over week.
- p50, p95 and p99 with a sample count
- Weekly trend: got worse or better, and by how much
- CSV export of the checks for external analysis
- Weekly report by email, with a downloadable PDF
Comparison
Sentinela vs alternatives.
| Sentinela | Free international monitor | Paid international monitor | Observability platform | |
|---|---|---|---|---|
| Minimum interval | 1 min | 5 min (free) | 1 min | 30 s |
| Native Portuguese | yes | — | — | — |
| Payment in BRL with invoice | yes | — | — | — |
| Hosted in Brazil | yes | — | — | — |
| Includes security auditing | yes | — | — | — |
| Public status page | yes | yes | yes | yes |
Every product is strong at something — pick the one that fits your case. Comparison based on the public plans of the respective platforms in August 2026.
For technical teams
API, webhook, isolation — as expected.
REST API
Auth via Sanctum. Read monitors, checks and incidents; write from Business up.
Webhook with retry
Versioned payload, custom headers, exponential retry until confirmation.
Native multi-tenant
Global scope in the database: each user only sees what is theirs, guaranteed in the query.
2FA TOTP
On every plan. Compatible with Google Authenticator, 1Password and Authy.
Internal audit
Trail of login, lockout and monitor/channel CRUD — exportable on Business.
CSV export
Check history for each monitor in CSV, via the interface or a dedicated endpoint.
How long until an outage is detected?
The interval you configure, starting at 1 minute. The outage only becomes an incident after being confirmed on consecutive checks — and the alert goes out seconds after that. That is what keeps a network hiccup from waking you up.
Where do you check from?
From our infrastructure in Brazil. It is the same network path your Brazilian customer takes, so the latency we measure is the one they actually feel.
Can I monitor a service behind a VPN or firewall?
Yes, with Heartbeat mode: your job pings a unique URL of ours every cycle. If the ping does not arrive within the expected interval, we go DOWN. Nothing needs to be exposed to the internet.
Does a high volume of webhooks cost extra?
No. Webhook, email and the other channels are unlimited on every paid plan. We charge per monitor, never per event — it makes no sense for you to pay more on the day everything goes wrong.
Can I tune how many failures become an incident?
Yes. You define how many consecutive checks must fail to open the incident and how many must recover to close it, monitor by monitor.
How does the Vercel monitor work?
We query the official Vercel API on every check. A failed production deploy becomes DOWN; a runtime error becomes DOWN once it passes the threshold you set — zero means "tell me about any new error".
What is the Synthetic monitor?
A real browser runs your user flow: opens, fills, clicks, waits and validates text or URL, with encrypted credentials when a login is needed. It catches the gap between "responds 200" and "the app works". Available from Pro up.
Uptime and security in one panel — and risk becomes a number.
The real incidents from your uptime feed the financial risk calculation: observed downtime cost × incident probability. It only makes sense with both things in the same place.
See Security Audit →Exposure = observed downtime cost × incident probability × impact.
Start today
1 monitor, 5-minute interval, free.
No card, no deadline. You move up to the 1-minute interval and the eight modes when you need to.