RELEASE — 2026-10-03

Server monitoring for the servers you actually have

Most organisations do not run a thousand servers. They run eight, or thirty, or sixty: a domain controller, a file server, two application servers, a database, a few Linux machines nobody wants to touch. MeshServerWatch 1.0.0, released on 2 October, is a server monitor sized for that — an agent on each server, a Manager you run yourself, and a bill that is the number of servers times five dollars.

The two usual options

The first is the hosted agent. It sends everything to a cloud — every process, every log line, every user name inside those log lines — and you get a dashboard back. What you have given up is knowing where your servers' data is.

The second kind is the enterprise suite: licensed per node, quoted from a hundred nodes, delivered with a consultant. For eight servers that is not a monitoring purchase; it is a project.

What MeshServerWatch does

Two programs.

The Agent runs as a service on each Windows Server (2016, 2019 or 2022) or Linux machine (Ubuntu 20.04+, Debian 11+, RHEL / Rocky / Alma 8+, amd64 and arm64). Every ten minutes it sends one report: CPU average and peak, memory, every real disk, uptime, and on Linux the load averages. Error-level events — the Windows System and Application logs at Critical and Error, journald at err and above — go out within about thirty seconds as one-line summaries. If the Manager is unreachable, it queues six hours of metrics on disk and sends them later.

The Manager is a single binary that runs as a Windows service or a systemd unit on one server that stays on. It receives the reports on one port, HTTPS 8455, keeps thirty days of charts per server, decides when a threshold has been crossed, and emails you once when it has and once more when it clears. Defaults: disk 90 % warning and 95 % critical; CPU or memory at 90 % for two reports in a row; no report for 2.5 × the interval makes the server silent — silent, not down, because the Manager only knows that nothing arrived.

HTTPS is on from the first start: the Manager generates its own certificate, each agent is installed with its SHA-256 fingerprint and aborts the TLS handshake, before the token is sent, if the certificate differs. No trust-on-first-use, no switch to turn the check off. Tokens are stored on the Manager only as hashes.

What it deliberately does not collect

The agent sends numbers and one-line summaries, and nothing else:

• For each error event: the log, the source, the event ID, the time, and a summary of at most 300 characters — the first line of the message.
• Never the full message text, the event XML, user SIDs, user names, process lists or command lines.
• Not the Security log. Not Warning or Information events.
• The machine identifier is a hash of the Windows MachineGuid or of /etc/machine-id; the raw value never leaves the server.

The exact fields are in the guide and in the privacy policy. The Manager makes no outbound connection unless you configure SMTP, connect it to MeshWatch Central, or switch on the once-a-day update check, which is off by default. The licence key is verified against a public key built into the Manager, so it works on a network with no route out.

Install in three commands

On the server that will be the Manager (Debian and Ubuntu shown; on RHEL, Rocky and Alma use dnf and the .rpm):

sudo apt install ./meshserverwatch_1.0.0_amd64.deb && sudo meshserverwatch install --accept-eula

That prints the console address and the fingerprint. Create the first administrator in the console and issue a token on the Agents tab — or from the same terminal:

sudo meshserverwatch token issue --label "Head office"

Then, on each server to be watched, the line the token box prints:

sudo apt install ./meshserverwatch-agent_1.0.0_amd64.deb && sudo meshserverwatch-agent install --manager https://manager.example.com:8455 --token msw_... --fingerprint AB:CD:... --accept-eula

Within a minute the server appears on the Servers tab. On Windows Server each program has a setup wizard; the Agent's asks for the same three values and has a Test connection button. For many servers the same setup runs silently from a GPO or an RMM tool:

MeshServerWatch-Agent-1.0.0-setup.exe /S /ACCEPT_EULA=1 /MANAGER=https://<manager>:8455 /TOKEN=<msw_…> /FINGERPRINT=<sha256>

The Windows installers are Authenticode-signed; the certificate is new, so SmartScreen may warn until it has built reputation. They were run through on Windows Server 2019 during release testing.

Pricing

US$5 per server per month, or US$48 per server per year, with a minimum of three servers. The Manager is included. Every Manager starts with a 30-day trial with unlimited servers, counted from the first agent report — no card, no account, no email address. There is no free tier after the trial. Charges are generally non-refundable, except where the law requires a refund or where we approve a refund request for a billing error submitted within 14 days of the charge (see the Terms of Use), so please evaluate during the trial. A server that is already reporting is never cut off when the seats run out; only a new server beyond the seats is refused, and the console says which one.

What is not in 1.0

• No MSI packages yet. The setup wizards install silently with /S, which covers GPO and RMM deployment; an MSI package is planned for a later release.
• No Docker image for the Manager yet. It is one binary with a .deb and an .rpm; an image is planned for a later release.
• No macOS agent. MeshServerWatch watches servers.
• Forwarding to MeshWatch Central is optional and needs Central 1.3.9 or later, which ships on 13 October 2026. The connection is part of the plan; nothing extra is charged for it.

Where to go next

Downloads, checksums and the FAQ are on the product page; the guide covers tokens, thresholds, email, licence seats and the Central connection. Questions: support@meshwatch.app.