Version 1.0 · Last updated 20 June 2026

WP Maintenance Agent keeps your WordPress sites updated, monitored, and safe — without the white-knuckle moment of clicking “update” and hoping nothing breaks.

It works in two parts. A free plugin lives on each of your sites and acts as the eyes and hands. A secure app is the brain: it reads each site, works out a safe order to apply updates, makes the changes one at a time while checking the site stays healthy, and rolls anything back the moment it misbehaves. Then it sends you a clean report.

The point isn’t blind automation. It’s an assistant that thinks before it touches your site — and always takes a backup first.

This guide walks you from zero to your first maintenance run, then tours everything the platform can do. You can read it top to bottom, or jump to a section from the contents.

Getting started

You’ll be up and running in three short steps. No command line, no server access — just your WordPress login.

Step 1 — Install the free plugin

In your site’s WordPress dashboard, go to Plugins → Add New and search for Maintenance Agent (by Kreiswolke). Click Install, then Activate.

Not finding the plugin? Download it here

That’s the whole install. The plugin is free, and it’s useful on its own — but the real value comes when you connect it to the app in the next step.

Step 2 — Connect your site

Open the plugin’s Connect screen (you’ll find it in your WordPress menu) and click Connect. Your browser opens WP Maintenance Agent, where you’ll:

  • Create your free account (or sign in, if you already have one).
  • Review and accept the Terms.
  • Click Authorize.

That’s it — your site is linked. Behind the scenes, the plugin and the app pair using secure keys, and the app creates its own limited, revocable access to your site.

You never type your WordPress admin password into us. That’s deliberate: it’s safer for you, and you can cut the connection from your own WordPress at any time.

Step 3 — Your first scan and maintenance run

As soon as you connect, the app runs a read-only scan. It reads your site’s software, versions, and health, and gives you a plain-language summary of what it found — without changing a thing. It’s a risk-free first look at how the agent “thinks.”

When you’re ready, click Run now to start your first real maintenance run, or set a schedule so it happens automatically (see Automatic maintenance below). On a brand-new site, you can also turn on Step-by-step mode to watch the agent pause between each stage.

How it works

A quick mental model so the rest of the guide makes sense.

The plugin and the app

The plugin sits on your WordPress site. It reports what’s installed, takes backups, applies updates, runs security hardening, and can put up a “maintenance” or “coming soon” screen — all on command.

The app is where you log in. It holds your whole fleet of sites, runs the maintenance agent, schedules work, watches your sites’ uptime, and keeps your reports and history.

The plugin only ever acts on instructions from the app (or from you). It doesn’t store your notes, history, or settings — those live safely in the app.

Reasoning before acting

Every maintenance run follows a careful loop:

  1. Back up first — the agent takes a full database snapshot and copies each plugin, theme, and core folder before touching it.
  2. Read and plan — it reads your site’s software, your notes, and what it has learned from past runs, then works out a safe order (for example, updating a base plugin before its premium add-on).
  3. Update one item at a time — after each update it clears caches and checks your site is still healthy.
  4. Roll back if needed — if an update or a health check fails, the agent decides what to do: retry, wait, skip, or restore that one item — without unwinding the work that already succeeded.
  5. Report — you get a clear before-and-after summary.

Because it verifies every change against what’s actually on disk, a dropped connection mid-update never leaves it guessing.

Risk tiers

Every site has a risk tier that tells the agent how cautious to be:

  • Standard — straightforward sites; the agent runs and reports at the end.
  • Caution — sites with things like WooCommerce or complex caching; the agent asks for your confirmation before risky steps.
  • High risk — sensitive or fragile sites; the agent stops for manual review.

You set the tier per site, and you can change it anytime.

Your dashboard

Everything starts from your dashboard — your single view of every connected site.

The all-sites view

Your sites are listed with their health, pending updates, last run, next scheduled run, and uptime at a glance. Colour-coded status tells you which sites are happy and which need attention — and each “needs attention” flag links straight to the reason.

When you manage more than one site, you can filter and group by client, by tags, or by health, and refresh one site or all of them with a click.

A single site’s overview

Click any site to open its detail view, organised into tabs: Overview, Hardening, Stack, Schedule, Uptime, Time Machine, Reports, Notes, and a live view while a run is in progress.

The Overview tab is your at-a-glance card: health, open alerts, SSL and domain status, owner, risk tier, and tags.

Owners, tags, and notes

A few small touches make a fleet manageable:

  • Owner / client — assign each site to a client so reports go to the right place.
  • Tags — label sites however you like (for example woocommerce, vip, monthly) and filter by them.
  • Notes — jot down anything the agent should know about a site. Your notes are fed into the agent’s reasoning on every run, so it gets smarter about each site over time.

Maintenance runs

The core of the platform: keeping a site updated, safely.

Running maintenance yourself

On any site, click Run now. The agent takes a backup, plans the updates, applies them one at a time with health checks, and reports back. You can pause or cancel cleanly between steps, and watch the whole thing live.

Letting the agent do it

Prefer hands-off? Set a schedule (see the next section) and the same careful run happens automatically on the days and time you choose.

What the agent does, in plain English

While a run is live, you’ll see the agent narrate its work:

  • It estimates your site size and takes a backup, showing progress as it goes.
  • It reads your site and explains its plan before changing anything.
  • It updates each item, then clears caches and checks your site is still healthy — telling you the result for each one.
  • If something looks wrong, it decides what to do and explains why.
  • It finishes with a verdict and a summary.

Health checks and automatic rollback

After each update, the agent loads your site and confirms it’s working — not just the homepage, but the WordPress backend too, so a “coming soon” screen or a cache layer doesn’t fool it.

If a site genuinely breaks, the agent restores that item from the backup it took, re-checks, and (if needed) stops and flags it for you. Your earlier successful updates are kept. Nothing is left half-applied.

Step-by-step mode

For a new or delicate site, turn on Step-by-step mode. The agent pauses between stages so you can review and approve as it goes. It’s off by default and asks for confirmation when you switch it on, so it’s never enabled by accident.

Automatic maintenance

Set it once and let it run. This is the autopilot.

Setting a schedule

On a site’s Schedule tab, pick the days (daily, weekdays, twice a week, or a custom set) and a time of day. Schedules use your account’s time zone and handle daylight-saving changes for you. Each site has its own schedule.

What happens on a scheduled run

A scheduled run is the same careful, backup-first process as a manual one — with a few extra safety checks suited to unattended work. For example, it won’t apply an update that needs a newer version of PHP than your site runs, and it retries briefly through the kind of momentary hosting blip that would otherwise stop a run.

Monitoring and alerts

Maintenance is only half the job — knowing the moment something’s wrong is the other half.

Uptime monitoring

We check your site from outside every few minutes. If it goes down, we don’t just say “site down at 14:32” — we pull live detail from the plugin (database reachable? recent errors? a plugin auto-disabled?) and tell you how it’s down, with a link straight to the run that may be related and the rollback button.

We only raise an alert after a second confirming check, so a one-off blip doesn’t spam you. And if a site is intentionally in maintenance or coming soon mode, we recognise that and label it calmly instead of crying outage.

Each site’s Uptime tab shows your uptime percentage, average response time, and a history chart with any incidents marked.

SSL and domain expiry

We watch two deadlines that quietly cause outages:

  • SSL certificate — we read your live certificate and warn you before it expires. Auto-renewing certificates (like Let’s Encrypt) stay quiet until they really need attention.
  • Domain registration — we check your domain’s expiry and remind you well ahead (90, 40, and 7 days out). For domains whose registries hide the date (such as some country domains), you can enter the date by hand.

When a certificate or domain renews, the related warnings clear themselves.

Daily health checks

Once a day, each site gets an automatic check of its software inventory, HTTP response, database, and error log. This surfaces problems on your dashboard before they cause a failed run. You can pick the hour per site, or use the default.

The emails you get

We aim for “useful, never noisy”:

  • Urgent things (a site down, a certificate expired, a domain due in 7 days) arrive immediately.
  • Non-urgent things (a certificate due in 30 days, routine notices) roll up into a single daily digest rather than a stream of separate emails.

Site tools

Beyond maintenance, the plugin gives you hands-on tools for each site — all driven from the app, so you rarely need to log into WordPress itself.

Security hardening

A full hardening layer you control per site, from the Hardening tab. Every option is off by default; you switch on exactly what you want, and each comes with a plain explanation of what it does and any risk.

Highlights include a login gate (a PIN page in front of wp-login.php to keep bots out), blocking common attack paths (like XML-RPC and author enumeration), security headers, and hiding usernames. Managing a fleet? You can harden many sites at once from the sites list.

There’s a master kill-switch that disables all hardening instantly if a rule ever gets in your way, and every change to your site’s .htaccess is backed up first so it’s always reversible.

Coming Soon and Maintenance mode

Two front-end screens you can flip remotely:

  • Maintenance mode — a temporary “we’ll be right back” splash, used automatically during risky updates and toggleable by hand. It never locks the app out of your site.
  • Coming Soon — a launch screen for sites that aren’t public yet, with an option to hide the site from search engines.

Both are recognised by our uptime monitor, so putting up a splash never triggers a false “your site is down” alert.

The cross-site installer

The Installer is a portfolio-wide plugin and theme manager. Browse the WordPress.org catalogue alongside your own fleet’s inventory, pick what to install, activate, update, or remove, choose which sites, and apply — to many sites at once.

By default each site is wrapped in the same safety net as a maintenance run (backup → apply → health check → auto-rollback if unhealthy), and you see live progress per site as it works.

Backups and restore points

Safety is built into everything above, but it’s worth understanding on its own.

How rollback works

Before touching anything, the agent copies each plugin, theme, and core folder and takes a full database snapshot, stored right on your own server. If an update goes wrong, restoring is a fast, surgical copy-back of just the affected item — no giant archive to unpack, and no dependence on a separate backup plugin.

Every restore re-reads the version from disk afterwards to confirm it actually landed.

Restore points and history (Time Machine)

Successful runs leave a snapshot you can return to. The Time Machine tab is a single timeline of every run and every saved snapshot — newest first, each with its date, result, and a one-line summary. Expand any entry to read the full report, or restore a past snapshot in one click. How many restore points we keep depends on your plan.

Emergency restore

A companion safety component can recover a site even when the main plugin is broken or disabled — the kind of “the site won’t load at all” situation where ordinary tools are useless.

The AI assistant and connecting your own AI

You can drive WP Maintenance Agent by clicking — or by talking to it.

The in-app assistant

Open the chat panel and type what you want in plain language, like “which sites have updates?” or “run maintenance on the bakery site.” The assistant works out what you mean and shows you a card to confirm. Reading is instant; anything that changes a site always asks you to confirm first.

Connecting your own AI (MCP)

You can connect your own AI assistant — such as Claude on the web, desktop, or your phone — to WP Maintenance Agent, and manage your fleet by chatting with it. Connecting is a one-click, secure authorization (no copying tokens around), and you choose what it can do: read-only, or read-and-run. You can revoke access instantly at any time.

This is genuinely unusual: maintaining your self-hosted WordPress fleet from your own personal AI assistant. It only works because the safety model underneath it is strict.

How we keep it safe

Whether a request comes from a button, the in-app chat, or an outside AI, it goes through the same protected core:

  • An AI can only trigger a fixed list of safe actions — never free-form commands, never raw database access.
  • Reads are free; every change requires a separate human confirmation — a single stray or hallucinated message can’t alter your sites.
  • Destructive actions (like deleting a plugin) aren’t available to chat at all.
  • Every action is logged, and if your assistant ever starts a run, you get an email heads-up — so an action you didn’t take is an early warning to revoke access.

Reports

Keep clients informed without writing a word.

Client and agency reports

Each run can produce two flavours of report: a technical version for you, with the full result table, and a plain-language version for your client, free of jargon. Reports are available as polished HTML, and you can add your own intro and sign-off so they sound like you.

Setting up delivery

Choose how often reports go out — every run, every few runs, or off — and set it separately for you and for your clients. Assign each site’s client (or use your client address book), and configure your sending details. There’s a master switch to pause all client email at once if you ever need to.

Plans and limits

WP Maintenance Agent is free to start, with paid plans that add automation, more sites, and agency tools. Exact prices are on the pricing page; here’s what each tier is for.

What each plan unlocks

  • Free — one site, with manual maintenance, basic monitoring (daily health checks plus SSL and domain expiry alerts), full security hardening, and a handful of agent runs to try it out. Great for a single personal site.
  • Solo — adds the autopilot: scheduled automatic maintenance, continuous uptime monitoring, reports, the in-app assistant, and connecting your own AI. Still one site.
  • Studio — everything in Solo across several sites, with fleet tools like the cross-site installer and more restore points.
  • Pro — a larger fleet plus agency tooling: a client address book, per-site report routing, and separate client and agency reports.
  • Agency — the biggest fleet, plus white-label branding on client reports and priority support.
  • Enterprise — custom limits and arrangements; get in touch.

Run limits and site counts

Each plan includes a number of agent runs and a maximum number of sites. Manual tools stay available within your plan; when you reach a limit, the app tells you and points you to the upgrade. You can change plans at any time from your account page.

Help and troubleshooting

Common questions

Do I have to give you my admin password?
No. When you connect a site, the app creates its own limited, revocable access — your admin password is never sent to us.

What happens if an update breaks my site?
The agent took a backup first. It restores the affected item automatically, re-checks the site, and flags the issue for you. Your other successful updates are kept.

Is the plugin useful without a paid plan?
Yes. The free plugin reports your site’s software and health and provides backup and hardening tools. The paid app adds the automation, scheduling, monitoring, multi-site dashboard, and reporting.

Can I disconnect a site?
Anytime — from the app, or by removing the access from your own WordPress. Uninstalling the plugin also cleanly ends the connection.

Why didn’t my site show as “down” during maintenance?
Because we recognise your maintenance and coming-soon screens and label them calmly, rather than raising a false outage.

How do I connect my own AI assistant?
From your account settings, use the connect option to authorize an external assistant over MCP, and choose read-only or read-and-run access. You can revoke it instantly.

Getting in touch

Need a hand? Email support@kreiswolke.com and we’ll help you out.


WP Maintenance Agent is a service of Kreiswolke, a brand of Anders und Anders Digipart GbR. See the Terms of Service and Privacy Policy for the legal details.