● uptime monitoring / single binary

MONITRA_

One binary.
Zero infrastructure.

Copy it to a box, run it, get alerted. No Docker, no database server, no config-file archaeology.

7.0 MB · HTTP / TCP / ICMP / K8s / host agents · CLI · TUI · web · API

~ shellbash

$ monitra monitor add --name api \ --target https://api.example.com --kind http --interval 30 $ monitra start ✓ monitoring started $ monitra tui # in another terminal _

## 01 the-gap

Two ways to monitor. Both cost you something.

--- hosted-saas Pingdom, UptimeRobot, Better Stack

✓ Great UX.
✗ Someone else's infrastructure.
✗ Cost scales per monitor.
✗ Private networks need an agent.

--- self-hosted Uptime Kuma, Zabbix, Prometheus Blackbox

✓ Powerful.
✗ Containers, databases, reverse proxies, config files.
✗ Real setup effort between “is it down?” and “I'm alerted.”

+++ monitraMonitra is the whole platform, in one executable, with no external service dependencies for its default setup.

## 02 proof

Numbers, not adjectives.

Measured 2026-10-01 on a 2 vCPU / 512 MB cgroup. Read the benchmark results.

benchmarks.txt

$ cat benchmarks.txt

METRICVALUENOTE

binary_size7.0 MBstripped release binary, default features

monitors1,000at 30s intervals, p99 drift 3× under the 2 s bound

scaling_limit~2,500CPU-bound; passes by ~1%, fails at 5,000. Single run, mock targets share the CPUs

p99_drift632 msN = 1,000 (1,979 ms at N = 2,500)

missed_checks0N ≤ 5,000 in the constrained run

static_musl6.8 MBno libc dependency; runs on Alpine or scratch

rss_at_500041 MBmemory is not the constraint

## 03 probes

HTTP. TCP. ICMP. Kubernetes. Hosts.

● http

Request an endpoint. Check the response.

● tcp

Open a connection to host:port.

● icmp

Ping the host.

● kubernetes

Deployments, StatefulSets, Services. Polled directly via the cluster API.

● host-agent

A lightweight agent pushes host checks back to the daemon.

Runs where your services are. Inside the VPC, on the edge device, in the cluster. A lost agent or broken collector shows as stale or unknown, never as down.k8s: direct poll · hosts: agent push

## 04 interfaces

CLI first. Everything else is a client.

cli

$ monitra <command>

Every capability. The reference interface.

tui

$ monitra tui

Terminal UI over SSH.

web

embedded, optional

A dashboard served by the same binary.

api

REST + WebSocket

For scripts and your own tooling.

> Every capability is reachable from the CLI. The web UI is a client, not a requirement.

## 05 alerts

Get told. Without blocking a probe.

log

default

Needs nothing external. Works on a pristine machine.

webhook

webhook://host/path

JSON message over HTTPS.

slack

slack://host/path

Slack text payload. Built with the slack Cargo feature.

> Bounded queue, 30 s retries, failures logged loudly. monitra alert list keeps the history even when delivery fails.

## 06 scope

What Monitra is not.

Scope discipline matters more than feature count.

NOT THIS

GO HERE INSTEAD

✗ Metrics / observability platform

Prometheus, VictoriaMetrics

✗ APM / tracing

Jaeger, Datadog

✗ Log aggregation

Loki, Elasticsearch

✗ On-call / alert routing

Alertmanager, PagerDuty

We say no to a lot so the binary stays small and the failure modes stay boring.

## 07 reliability

If we're uncertain whether a target is down, we say unknown. We don't guess.

A monitor that lies is worse than no monitor.

✓ up✗ down? unknown

## 08 get-started

Get from source to your first check.

Early software: v0.0.x. Build from source today; prebuilt binaries are not published yet. Known gaps are listed under limitations.

~ shell

$ git clone https://github.com/rustiqz/monitra.git $ cd monitra $ cargo build --release $ ./target/release/monitra monitor add \ --name api --target https://api.example.com \ --kind http --interval 30 $ ./target/release/monitra start _

[ Releases ]changelog →[ Read the docs ]→Source on GitHub→