Get early access
Features · Monitoring · Metrics & Charts
Monitoring · Metrics & Charts

Per-second telemetry,
no exporters to install.

CPU, memory, disk, load, network, processes — collected by the agent every second, charted with proper p50 / p95 lines, kept for two years. Sparklines on every list, full charts on every host.

web-prod-04 · 1s resolution
cpu 18% p95 41% steal 0.2%
mem 33% 12.1 / 32 GB cache 6.4 GB
disk / 71% /var 44% inodes 12%
load 0.84 0.61 0.55
net rx 41 Mb/s tx 12 Mb/s retrans 0
retained 2 years · 1 sample / second
— Why this matters

Most "monitoring" is exporters, dashboards and glue.

01

One agent, not one per service.

The metrics arrive because the agent is installed, not because someone wired up a node exporter and a scrape config.

02

Per-core and per-device, not just averages.

An average CPU of 40% hides one core pinned at 100%. A total disk figure hides the one device with a queue.

03

Retention should not be a cliff.

Raw samples for the recent window, hourly aggregates beyond it, daily kept indefinitely — so last year is still answerable.

— What's included

What the agent actually collects.

CPU
Overall and per core, split into user, system, iowait, steal, irq and softirq — steal being the one that tells you it is not your fault.
Memory
Used, cached, buffers, swap, available, slab and HugePages, separated rather than summed into one misleading number.
Disk
Usage and inodes per mount, plus per-device IOPS, throughput, latency and queue depth.
Network
Bytes, packets, errors and drops per interface, not just the aggregate.
Processes
State counts including zombies, with the top five by CPU and by memory.
TCP connections
A state histogram — ESTABLISHED, TIME_WAIT, LISTEN, CLOSE_WAIT — which is where connection-leak questions get answered.
File descriptors
Open against the system maximum.
Load and uptime
One, five and fifteen minute averages.
Charts with range
1H, 6H, 24H, 7D, 30D, sparklines in lists and full charts on the host.
— What it looks like

CPU on one host, 1-second resolution.

23% alert 85%
Example view · web-prod-04 · last 24 h · sampled every second, rolled up for display
— Works with

Connected capabilities

Service Plugins
Deep metrics for the services the agent discovered.
→
Custom Dashboards
Compose your own views from the same metric pool.
→
Capacity Planning
Forecasts derived from the same time series.
→
Smart Alerts
Build alert rules on the metrics that already exist.
→
— FAQ

Questions, answered honestly.

How much does collection cost the host? +
Negligible at the cadences the agent uses by default. We measure overhead on every release; current numbers live in the agent README.
How long is data kept? +
Two years at native resolution on every plan. After that, downsampled to 5-minute resolution and kept indefinitely.
Can I export raw metrics? +
Yes — Prometheus remote-write format from the platform, or per-host scrape endpoint if you prefer pull.

Charts that don't lie.

Free tier includes 7 days of native-res data; longer retention from Starter.

Get early access Back to Monitoring