Peek is open-source server fleet monitoring with a CLI-native dashboard. One command adds any Linux server — real-time telemetry, historical graphs, and zero ops overhead.
No agents to babysit. No dashboards to build. SSH in once, never look back.
A clean, CLI-native interface. No bloat, no configuration, no dashboards to build. Just your servers and their metrics, live.
Meaningful telemetry with no configuration. Everything you actually need, nothing you don't.
Self-configuring installer detects Python, installs psutil, registers the node, and starts a systemd service. No config files, no restarts, no manual steps.
CPU per-core, load averages, memory, swap, disk I/O, network throughput, process table, temperatures, uptime, and hardware — not just the highlights.
Live charts with time windows from 5 minutes to 30 days. Rollup retention keeps storage lean while keeping enough history to actually investigate an incident.
Total nodes, aggregate vCPU, memory, and disk across your entire fleet at a glance. Spot which server is the outlier without clicking through each one.
One-time or reusable tokens with device caps and expiry. Give a token to a teammate, revoke it the moment it's no longer needed. No shared credentials.
Nodes crossing 95% CPU or memory are surfaced before you even think to look. Catch runaway processes before your users notice anything is wrong.
The agent pushes data outbound over HTTPS. No ports to open on the server side, no VPN, no static IPs. Works behind NAT and in cloud private networks.
The dashboard updates via Server-Sent Events. No page refresh, no client-side polling — the server streams changes to you as they arrive from the agent.
Self-host the entire stack yourself, or use this instance. The agent, dashboard, and hosted multi-tenant layer are all open source and on GitHub.
Features that only make sense when Peek runs the infrastructure for you. Sign in and they are already there.
Invite teammates by GitHub username. Owner / admin / viewer roles control who can add servers, manage tokens, or read the fleet without modifying it.
Share a read-only status URL for your workspace — no login required. Includes per-node graphs. Embed it in your runbook or send it to a customer.
Per-workspace data volume tracking so you know which team generates the most telemetry before hitting a retention limit.
Daily or weekly fleet health summary in your inbox — top CPU consumers, disk pressure trends, and nodes that went offline overnight.
30-day per-node availability history. Set SLO targets and track incidents without running a separate uptime tool just for this.
"At current write rate, /var fills in 6 days." Proactive headroom warnings based on your actual trend, not static percentage thresholds.
Track TLS cert expiry for domains served by your nodes. Alerts at 30, 14, and 7 days out — before the browser starts showing security warnings.
Statistical baseline per metric per node. Alerts fire when usage is significantly above normal behaviour, not just when it crosses a blunt static threshold.
When Docker is present, break down CPU and memory by container — so you know which service is burning resources, not just which host.
Fire a JSON payload to any URL on threshold events. Connect to PagerDuty, Discord, ntfy, or your own backend with no code on your side.
Each cron job pings a unique URL on completion. If the ping doesn't arrive in the expected window, you get alerted — catches silent cron failures.
Mark a deploy on the timeline and see exactly what happened to CPU, memory, and error rate after it landed. Correlation becomes obvious.
Any Linux server. Root access. That is genuinely all you need.
Your workspace is created automatically on first sign-in. Servers, tokens, and telemetry are scoped to your account from the start.
Open Tokens and generate a one-time or reusable token. The install command is pre-filled — just copy and paste it.
Paste on any Linux host as root. The node appears in your dashboard in seconds.
Every feature. Unlimited servers. No credit card, no trial, no upsell.
No trials. No tiers. No surprises. Ever.
Self-host it yourself or use this instance. The code is all there, including the agent, the dashboard, and the hosted multi-tenant layer. Pull requests are welcome.