NimBox SRE is the operation layer that turns your alerts into billable service: an agent with auto-discovery, native PAM and eBPF profiling that operates hosts after NAT, without opening SSH. Incidents, approved runbooks and monthly evidence ready for your SLA. Without mounting a 24×7 NOC.
You have 50–500 hosts divided between SMEs, a contained team and clients who pay maintenance or SLA. The problem is not detecting: it is operate, control and demonstrate what your team does.
You connect what you already have, you define how to act and each incident leaves evidence. Your team operates with context; you invoice with proof.
Monit (native), OpenTelemetry and Prometheus fall into a single operational layer. Context by host: status, metrics, processes and incident history.
// don't set up another siloA single command: download the binary, auto-detect the running services (nginx, postgres, docker, redis...), register the host in PAM and start reporting. Without configuring anything, without opening incoming ports — it works the same after NAT.
// from zero to operand in 60 secondsEach shutdown leaves ownership, severity, session, time, and SLA compliance. The monthly report is generated on its own, it is not rebuilt.
// service that is explainedThis is the life of a host from the moment you install it until its incident is closed and justified. No VPN, no exposed SSH, no touching the client's router.
A curl | sh with the client's token. The agent works out on its own what is running —containerd, cron, nginx, sshd—, registers itself in PAM, drops its systemd unit and starts reporting. You configure nothing. You open nothing.
By the first cycle the host is already there with its inventory resolved: distribution, kernel, provider, location, metrics and even the K3s node with its pods. Real plug & play: nobody registered anything by hand.
Docker, Kubernetes, Proxmox, Zimbra, Asterisk, Plesk, Redis, MySQL, TLS certificates, GPUs… If the software is on the machine, it gets watched. There is no screen to configure because there is nothing to decide — and you can answer “are you monitoring this?” without opening the code. 20 and growing…
You connect Telegram, Slack, Google Chat or WhatsApp and the agent lives where your team already talks. You ask about a host in plain language and it answers with data, not with a link to yet another panel.
The very first check already spots two systemd units down. The agent opens the incident, runs the approved runbook and leaves it resolved —with the detail of what failed and which command verifies it. And it tells you through whichever channel you picked.
People, agents and automations leave the same trail: who, what, on what and with what result. Every record is chained to the previous one —tamper with one and the chain breaks— and each intervention keeps the recording of what was run. This is what you show when the client asks.
Designed for MSPs, software houses, and infrastructure teams that need to turn complex operations into clear—and billable—results for their clients.
Discovers and watches whatever is there: Asterisk, Zimbra, Carbonio, Plesk, MySQL, Redis, PHP-FPM, Varnish, Docker and Swarm, certificates, SMART disks, GPUs. And what does not live on the host — AWS with CloudWatch, Proxmox, network quality towards whatever matters to you.
Ownership, severity, evidence and SLA to close the complete operational cycle.
Single binary of a file, without dependencies. It detects services, processes and containers only, operates behind NAT and activates eBPF profiling (I/O latency, kernel stacks) when the metric is not enough.
The agent self-registers in PAM upon installation. Every SSH connection is audited. No passwords in traffic, no manual configuration.
Turn activity, resolutions, and SLA compliance into a useful monthly report.
Opens and closes tickets in Jira or GitLab when an incident needs a person. Talks where you already talk: Telegram, Slack, Google Chat or WhatsApp — ask and reply in the same thread. And your metrics in Grafana — dashboards already built, or connect your own. No tool migration.
Converts each resolution into an approved runbook, versioned and linked to the affected host or service. The agent—or any technician—acts within a framework that your team reviews and controls.
The agent works for pull: It goes to NimBox via HTTPS and picks up what it has to execute. Runbooks run on the machine even if it's behind NAT, CGNAT, or a client home router — without exposing SSH, without opening a single incoming port, without tunnels to maintain.
The agent focuses on diagnosing and, using data—metrics, processes, load, and history in a single view per host—finds the source of the problem to truly resolve it. Fewer returning incidents, and happier customers.
Self-discovery + eBPF. The agent detects what is running on each host without configuring anything. And when the metric is not enough, activate eBPF profiling on the fly: I/O latency histogram and kernel stacks of blocked processes, with the exact time they spend stuck. It turns on only where the kernel supports it, without heavy agent or instrumenting anything.
NimBox SRE doesn't give you another panel: it gives you capacity. The same team closes more incidents, with controlled procedures and evidence that you can invoice every month.
More closures per technician thanks to runbooks and agents that do repetitive work.
Less time per incident: context per host and procedure at hand from the first minute.
First response and resolution with configurable objectives and visible compliance.
The agents cover the guard: they investigate and ask your permission before acting. Coverage without the cost of a real 24x7.
No. NimBox SRE comes with its own lightweight agent, but coexists with your existing monitoring- Natively integrates with Monit and supports OpenTelemetry (OTLP) and Prometheus, and turns those alerts into operation with context, evidence and SLA. You don't change tools or migrate anything — you put it next to what you already have.
Never without control. Agents only run approved and versioned runbooks, scoped by host or service, and all access goes through a PAM that records and audits every agent action in a session. You can revoke any procedure in one click.
The goal is to pilot a real client in days, not months: you connect hosts and alerts, define the first runbooks based on what you already solve by hand, and start accumulating evidence from the first watch.
Each incident leaves ownership, severity, time, session and SLA compliance. The monthly report is generated from that evidence rather than reconstructed by hand — you use it to justify hours, defend the SLA, and demonstrate the value of the service.
We start with a short demo about your case and a limited pilot. Write to us at [email protected] and we set up a call to see if it fits.
Tell us what you operate and we'll show you how NimBox SRE fits in a short demo.