A single lightweight agent reports the health of every server in your fleet every few seconds. CPU, memory, disk usage, network throughput, load average, process counts. Set thresholds, get emailed when something melts.
Install once via curl. The agent runs as a systemd service, pushes metrics every 10 seconds, and self-updates. Linux x86_64 and ARM supported (macOS dev builds available).
All standard host metrics: per-core CPU, RSS/cache memory, disk usage and IOPS per mount, network bytes in/out per interface, load average, top processes by CPU/RSS.
Set per-host or per-tag thresholds (e.g. 'alert if CPU > 90% for 5 min on tag=production'). Email on trigger and recovery, no PagerDuty contract required.
Server monitoring used to be a separate category — Nagios, Zabbix, Datadog, New Relic. The tools are powerful but the pricing models punish growth: $15/host/month becomes $1,500/month at 100 hosts, and that's before you add APM, RUM, or logs. Datadog's bill is a running joke in the engineering community for a reason.
Nevision bundles host metrics into the same flat-rate plan as session replay, errors, uptime, and logs. There's no per-host cost. Pro at $39/month covers a fleet of dozens. Business at $99/month covers a fleet of hundreds. If you're growing past that, talk to us about an Enterprise plan — it's still flat-rate, just with higher limits.
Every 10 seconds: per-core CPU usage, total memory used/free/cached/buffered, swap usage, load average (1/5/15 min), per-mount disk usage and free space, per-mount disk read/write bytes and IOPS, per-interface network bytes/packets in and out, TCP connection count, top 10 processes by CPU and RSS. We also collect uptime, kernel version, distro, and hardware metadata so dashboards can group by platform.
Host metrics are useless without grouping. Every host can carry arbitrary key=value tags. Tag by environment (env=prod), role (role=api), region (region=us-east-1), cluster, team, or anything else. Every chart, every dashboard, every alert can be scoped to a tag query (e.g. env=prod AND role=api). Add or remove tags via the agent CLI without restarting.
Track specific processes (e.g. nginx, postgres, your_app_binary) — alert when a named process disappears, crashes, or exceeds memory thresholds. Useful for catching silent crashes that your app's own healthcheck might miss.
One curl command. Sets up a systemd service, opens an outbound connection to api.nevision.app:443, no inbound ports needed.
Group servers with arbitrary key=value tags (env, role, region, cluster). Every chart and alert can be scoped to a tag set.
Build dashboards from the metric library, set alerts on any metric (e.g. memory.used_pct > 90 for 5m), receive email or webhook on trigger.
| Feature | Nevision | Datadog | New Relic | Netdata |
|---|---|---|---|---|
| Free hosts | Up to plan limit | 5 (limited) | 100 GB ingest | Unlimited (community) |
| Paid plan starts at | $12/mo | $15/host/mo | $0.30/GB ingest | $5/node/mo |
| CPU/memory/disk/network | ||||
| Includes session replay | Add-on | Add-on | — | |
| Includes uptime monitoring | Add-on | Add-on | — | |
| Includes error tracking | Add-on | — | ||
| Self-host option | — | — | — | |
| Per-host pricing trap | — | Usage-based | — |
Comparison based on publicly listed pricing and features as of April 2026.
Linux x86_64 and ARM64 (Ubuntu, Debian, RHEL, CentOS, Alpine, Amazon Linux). macOS for development. Windows is on the roadmap for 2026 H2.
Under 0.5% of one CPU on a typical host. Memory footprint is around 30 MB RSS. The agent is written in Go and statically compiled.
No. It opens an outbound HTTPS connection to api.nevision.app:443. No firewall changes needed for inbound traffic.
Container-level metrics (per-cgroup CPU, memory, network) are supported on hosts running Docker. Kubernetes-native deployment via DaemonSet ships in 2026 Q3.
Email by default. Webhooks (POST JSON to a URL of your choice) for paid plans. PagerDuty and Opsgenie integrations are on the roadmap.
Zero. Plans are flat-rate. Free covers 1 host (lightly), Pro covers up to 25, Business covers up to 100. Beyond that, contact us for Enterprise.