AI agentAI agents
Operations watch: an agent that checks what keeps your business running, and tells a person when it breaks
Our operations watch agent checks, around the clock, the things that quietly stop a business when they fail: websites and apps, servers, SSL certificates, backups, payment and lead forms, scheduled jobs. It sends a short morning brief, alerts the right person at once when something is down, and suggests the first step, but it only takes actions you have listed in advance, such as restarting a service, and it never deletes anything.
What we do for you
- An inventory of what must keep running: sites, apps, servers, domains, certificates, backups, forms, scheduled jobs
- Health checks every few minutes, with a status page your team can open
- Expiry watch for SSL certificates and domain names
- Backup checks: did last night's snapshot run, and can it be restored
- End-to-end checks of what earns money: a test lead through the form, the checkout page loading
- A morning brief: what failed, what recovered, what needs a decision
- Alerts by chat, email or phone to the person on duty, with the context
- A list of pre-approved actions (restart a service, roll back a release), everything else left to a person
Who it is for
- Fits: small businesses whose website, shop or app is their revenue, without an operations team
- Fits: agencies hosting many client sites who learn about outages from the clients
- Fits: teams with scheduled jobs (imports, emails, reports) that fail silently
- Does not fit: replacing a 24/7 security operations centre for regulated systems
The problem it removes
Things rarely break loudly. A certificate expires on a Sunday, a contact form stops sending, a nightly backup has been failing for weeks, a scheduled import silently stopped. The owner finds out from a customer, or when it is time to restore and there is nothing to restore.
Some of these failures have legal weight too. Under the GDPR, a personal data breach must be notified to the supervisory authority without undue delay and, where feasible, within 72 hours of becoming aware of it, and every breach must be documented with its facts, effects and remedial action (article 33). Knowing quickly that something went wrong is the first step.
How it works, step by step
- Inventory: we list what must keep running and who is responsible for each item.
- Checks: health checks on each site and service, certificate and domain expiry, backup status, end-to-end tests of forms and checkout.
- Status page: one page showing what is up, what is down and since when.
- Alerts: a failure alerts the person on duty, with the evidence and the likely first step.
- Approved actions only: the agent may restart a service or roll back a release if you listed that action; anything else waits for a person.
- Morning brief and weekly review: a short summary every morning, and recurring issues reviewed weekly.
What it does, and what it never does
| The agent does | The agent never does |
|---|---|
| Checks your services on a schedule and keeps the history | Deletes a server, a database, a file or a backup |
| Alerts the right person with the evidence | Hides or retries a failure silently without telling anyone |
| Takes only the actions you pre-approved | Changes DNS, payment settings or access rights on its own |
| Watches certificate and domain expiry dates | Renews or buys anything that costs money without your OK |
| Writes a morning brief | Contacts your customers about an incident |
What it plugs into
Most of what the agent needs is already there. Docker Compose, for instance, lets each service declare a health check (a test command run at an interval, with a timeout and a number of retries) and lets a service wait until another is reported healthy before starting. Let's Encrypt certificates are valid for 90 days by default and Let's Encrypt recommends renewing every 60 days, so expiry dates are worth watching even when renewal is automated. The agent also uses:
- Your servers and hosting accounts (read access, plus the actions you approve).
- Your website, shop and app addresses, for external checks.
- Your backup system, to check the last run and test restores.
- Your team chat, email or phone for alerts.
What we already run ourselves
We run this on our own infrastructure. Our servers are checked every 5 minutes and the results are shown on a health page; a morning brief and a night-ops review run every day; our server takes daily snapshots; our deployments have health checks and automatic rollback; and SSL renewal is automated, including on four servers we look after on AWS. The client version uses the same pieces, set up on your infrastructure and in your name.
Questions we get
Will the agent fix problems by itself?
Only with actions you have approved in advance, such as restarting a service or rolling back a release. Anything else, especially anything destructive or costly, waits for a person.
Who gets alerted at night?
The person you name for each item. Non-urgent issues wait for the morning brief; outages of what earns money alert at once.
Do you need full access to my servers?
Read access is enough for most checks. For approved actions, we use the narrowest access that allows them, in your name, and you can withdraw it at any time.
Does it help with GDPR breach obligations?
It helps you know quickly that something went wrong and keeps a log of what happened. Deciding whether an incident is a personal data breach and notifying it stays your responsibility, with your data protection adviser.
Sources
- Docker Docs: Compose file reference, services (healthcheck, depends_on) (checked 2026-10-06)
- Let's Encrypt: FAQ (certificate lifetime and renewal) (checked 2026-10-06)
- CNIL: GDPR text, chapter 4 (article 33, breach notification) (French) (checked 2026-10-06)