Repair
Fixes it, then proves the fix.
Xovris finds the cause and puts it right by itself, as far as the level you set allows.
- Readnorthmerefreight.com/checkout
- FoundDNS
- Fixed
198.51.100.66203.0.113.10 - Proved0 of 3 3 of 3
XovrisOn its own
- A lit square is a feature people use today
- Newest: Undo a fix, or let Xovris put it back itself
Every feature in Repair, strongest first.
What fixes and prevents a problem comes first. Type to find one, or open a plan to see only its own.
Nothing here matches those words yet. Xovris notes the words, with no address and no cookie, so what is missing gets written.Ask Xovris instead
Lit: live and workingOpen: Xovris is still working on it
Undo a fix, or let Xovris put it back itself
Every fix card has its change, its risk and one Approve; if the check after a change fails, Xovris restores the earlier version by itself.
Every step said before it runs
Xovris says each step it will take and whether it is ready or waiting on your go, and never calls a step done before it has run and been checked.
Done means a check that could fail has passed
Xovris says a job is done only after a check taken from your own request - one that could have failed - has passed.
A failed attempt is diagnosed before the next one
When an attempt fails, Xovris names its cause before it tries another way, so the same mistake is not made twice.
Break a sample site and watch it fixed
On the website one button breaks a sample site, clearly labelled as a sample, and you watch Xovris find the fault and fix it live.
The home page plays one real fix from start to finish
The home page plays a fix Xovris really made, step by step: the failing check, the cause it found, the change and the check passing again.
Every repair tracked to its verified end
Every repair runs as a checked, authorised action and is tracked step by step until it is verified, mitigated or undone.
A bad feature flag switched off for you
When a failure comes from a feature flag in your flag service, Xovris names the change, switches the flag off within what you allow, and proves it by re-running the failed check.
Your own library of procedures
Keep your team's own procedures in one library; each fix state says what it means once, beside its count.
Xovris never runs in circles
If an action starts repeating, going back and forth or running past its budget, Xovris changes approach, slows down or asks you - and never drops the fix.
Every AI step watched for runaway work
Each step Xovris takes is checked for loops and runaway cost before it runs.
Set a monthly risk budget for what Xovris does alone
Set, in plain words, how much risk each kind of action may take on its own each month, and see how much is spent.
Deploys guarded by your error budget
When a journey's error budget is spent, new features are frozen until it recovers, and a burning budget pages the right person.
Every user-facing fix shows a before and after
Each fix to something your users saw carries a masked replay of the broken screen and of the screen after the fix.
Every change keeps a verified rollback point
Every kind of change Xovris makes on your systems keeps a rollback point that is checked to work.
Undo anything Xovris did in the last 30 days
One press undoes every change Xovris made in the last 30 days, each from its own rollback point.
Every finished job carries a proof receipt
Each finished job shows what changed, before and after, with the receipt and the rollback point behind it.
A failover record outside your zones connects in one press
When a failover record is in a zone Xovris does not hold yet, the answer carries the button that connects it.
Failures rehearsed every night on a copy
Inside your quiet hours, Xovris plays the faults it can name on a copy of your setup and closes the weak points it finds, such as a missing check.
Nothing is done without the challenger's review
The database itself refuses to record a job as done unless a second, independent review of its method is on record.
No job ends in a failed state
Every kind of job - an incident's investigation included - ends done with its proof, or stays in progress with its next step. Never simply failed.
A stalled request moves to its next approach by itself
A request Xovris is holding is tried again at its next approach automatically, and never ends in a failed state.
Email delivery problems fixed before mail bounces
When bounces or spam complaints creep up, Xovris rehearses the fix on a copy of your provider's settings first, then makes it through one safety gate and tells you.
A filling disk is cleared before it is full
When a computer's disk heads for full, Xovris clears only what is safe to clear, such as caches, before it runs out, and tells you what it did.
A stalled job switches approach
A job that runs past its time or keeps hitting the same wait is caught and switched to another approach.
Roll back or redeploy on any host
Pointing production back to the last good release, or starting a deployment again, works on any host with an API - not only the five Xovris knew first.
Switch off a bad flag on any flag service
A feature flag changed just before a failure is found and switched off on any flag service with an API, not only the eight Xovris knew first.
A change that does not hold is undone by itself
If a value Xovris typed or a change it made does not read back as expected, it is put back automatically and checked against how things were before.
DNS changes rehearsed on real DNS servers first
A zone change is rehearsed on real DNS server software in Xovris's sandbox and read back before your records are touched.
Email fixes rehearsed with a real delivery first
A mail fix is rehearsed in Xovris's sandbox with a real message sent and read back before it is made for real.
Cloud changes rehearsed before they touch your account
A change to cloud storage is rehearsed first on stand-ins of AWS and Azure in Xovris's sandbox, and read back there before it is made for real.
Kubernetes changes rehearsed on real clusters first
A Kubernetes change is applied first to throwaway kind and k3d clusters in Xovris's sandbox, read back and undone there, before your cluster is touched.
Website fixes tested in three browsers first
A fix to a broken page is rehearsed in Chromium, Firefox and WebKit inside Xovris's sandbox before it reaches your site.
No fix runs until it is proven
A fix runs only once it is proven - in a copy of your setup first wherever one can be made - and a fix that failed its rehearsal never runs.
Planned work checked before it starts
Each change window in the next three days is checked first: an open incident or a failed backup puts the work on hold, with every reason and how long until it starts.
Missing security headers fixed in your code
When a site you have connected is missing a security header, Xovris finds the repository that deploys it, works out the change to its header file and makes it there, with a way back.
Xovris waits with you at a step only you can take
When a job needs something only you can do - a password, a security setting, a payment - Xovris holds the job there for as long as it takes and sends the system nothing that could make it worse.
A stopped job carries on by itself
A job that stopped at a wall is kept with its exact request and your yes, and Xovris carries it on from the same point once the way is clear.
Jobs that meet a wall find a way through
When a system answers a job with a wall - a sign-in, a code, a challenge, terms or a rate limit - Xovris recognises it and takes the way through instead of failing.
A planned job keeps going without you
A job Xovris planned carries on in the background from where it stopped, and is never called done until it really is.
A job that goes round in circles is caught
Every job has a watchdog. If it loops or goes quiet, Xovris switches to another approach and keeps a note of why the first one was dropped.
The right tool for a job Xovris has never done
When a job needs a tool Xovris does not have, it searches for one, checks every candidate and uses only one that passes.
Every place a tool could be is searched
Xovris looks in package registries, GitHub, the vendor's own tools, published AI-tool servers and APIs. A place that does not answer is noted as not answering, never as having nothing.
Free and open-source tools first
Of the tools that pass its checks, Xovris uses the best one, and chooses a free, open-source tool before a paid one.
One exact version, named on the job
Xovris uses one exact release of a tool, never whatever happens to be newest, and the job's record names it.
When no tool is good enough, Xovris builds one
If nothing that exists passes its checks, Xovris builds the tool itself and rehearses it before it touches anything of yours.
Every change is made safely, under your policy
Before any change Xovris shows exactly what it would send, the one reading that would move and the way back, rehearsed on a copy first; the real change follows once the rehearsal passes, under your policy. Name something you run through the API and Xovris answers with every reading it gives and keeps them for you. Xovris goes nowhere you have not authorised.
The results page opens on the fixes themselves: day by day, and one fix playing stage by stage
The results page no longer opens on two boxes of words. Its first screen is now the record: a strip with one mark for each day, lit only on a day a fix was proved, and under it one recorded fix as a line from detected to verified with the record’s own times. It plays once at the speed it ran, you can play it again, and with motion turned down the same picture stands complete and still. The fix’s report, what has to be true before Xovris says fixed, and where your own results live follow straight after it.
Your repository’s own scripts run sealed off, and Xovris finishes the work around a change – from a chat message to the follow-ups after the merge
Every linter config, build file, setup script and test command from your repository runs in its own sealed copy that holds no secret of yours or Xovris’s, reaches no network and starts nothing else, so a config that tries to read keys or phone home finds nothing. Ask in Slack, Teams, Discord or Google Chat and Xovris makes the change in that copy, runs your tests and answers in the thread with the pull request and its proof; a scheduled request works the same way. It resolves merge conflicts keeping both sides and proves the result with your tests, writes tests that are each shown to catch a real bug, adds documentation and simplifications only when your tests still pass and still catch a bug, and after a merge adds the changelog line, opens the issue, posts in your channel and watches production for the change.
Every fix says Verified, Not verified or Reverted – on the page, in the email and on your status page
No change is ever shown as done before it is proven. Every fix report, on the incident page, in the email and in Slack, says Verified with the check that now passes, Not verified with the reason, or Reverted with what was put back. Alert emails show the same state and proof, laid out as the app shows them, and a text message carries it too, inside one message. Your status page links each fix’s signed proof receipt beside the incident, with the key that checks it, so your own customers can verify what was fixed without asking anyone.
See what code review found and what your team did with it, week by week
Every review Xovris makes of a pull request is kept with its findings, and each finding settles by what really happened: fixed by Xovris with a proven fix, taken up by your team on a later commit, dismissed, or never merged. The new Code review results page, the API and your weekly letter show the same figures – raised, fixed, taken up, dismissed, still open and review time saved – each read from the stored reviews, and a quiet period says so instead of showing noughts. Code quoted from your repository is always marked as such, so nobody mistakes it for Xovris speaking.
Code review you can act on: accept a finding and Xovris fixes it
Accept any review finding and Xovris shows the exact patch first, then fixes it in a safe copy. Tell it how you like reviews in your own words, read a plain walkthrough of every change, work from one queue of open pull requests ordered by what each needs next, and filter findings on every axis at once.
Every change Xovris makes does what you asked and can be undone in one press
Before Xovris changes anything in one of your systems it reads the record first, rehearses the change on a copy of your system, and changes only the fields you named – a request that matches no record, or more than one, is stopped and explained rather than guessed. Afterwards it reads the result back and checks it field by field against your request, not against its own plan, so done means done; a reply that says success while the record says otherwise counts as a failure. Dates, amounts, units and time zones are understood the way you wrote them. Xovris asks you only when two readings of your request would end differently, confirms only what cannot be undone – deleting, sending, paying, publishing – and keeps a one-press undo for everything else. Every job can be replayed in order.
When a vendor breaks for one customer, Xovris protects everyone else first – running the safe fix ahead of time
When Xovris fixes a problem one customer met with a vendor – an issuer, a DNS provider, a host – every other customer using that vendor is protected before it reaches them. Only the fix travels, never the first customer's data. A fix that prevents the problem and can be undone, such as signing a zone or renewing a certificate early, runs ahead through Xovris's runner with its rehearsal, decision and undo. A fix that reacts to an outage runs only while the vendor's own public status page confirms it. Anything charged to your account waits for your yes, and everything else is held ready for the moment it is needed.
Xovris blocks a hostile address on your firewall itself, shown first and undone in one press
When the network watch sees an attack from outside, Xovris offers to block the address on your sensor or firewall (Suricata, FortiGate or Palo Alto): the exact change is shown first as a dry run, it is applied by Xovris on your yes, checked to prove it holds, ends by itself after the time you choose unless you keep it, and is undone in one press.
When your internet goes wrong, Xovris takes the next step itself: restarts the router, asks for a fresh address, writes to your provider and reads their status page
The network test names the part at fault, and now each next step is something Xovris does itself, within your policy: it restarts an OpenWrt router through its own settings and checks it, asks the router for a fresh address, tests the signal again after you move, sends your provider the case with every measurement and follows it up until it is fixed, and reads your provider's published status page itself. Each step still offers doing it yourself beside it.
Repository fixes, your badge and your free check go onto your pages as one pull request – Xovris writes them
When a fix is a change to your repository – the CA bundle in a minimal image, a liveness probe that waits on a database, container limits, a connection pool that drops idle connections, the JVM heap, ASP.NET Core's stdout log, rotating container logs or the dashboard panel on your page - pressing the step now does it. The same goes for your Xovris badge and the free check you embed with a partner key: Xovris no longer hands you the code to paste, it puts it on your pages itself. It reads only the files the change applies to, writes it on its own branch and opens one pull request through the Xovris GitHub app, after you consent once for that repository. Your default branch is never written, your own checks run before anyone merges, a second press finds the one already open, and the code stays beside it if you would rather place it yourself.
Review code on a Git server inside your own network, with nothing opened to the internet
You can now run the Xovris connector inside your network and Xovris reviews and fixes code on GitHub Enterprise Server, self-managed GitLab or Bitbucket Data Center without a single inbound port: the connector only ever calls out, keeps your Git server’s token inside your network, and refuses anything but that server’s review requests.
Xovris reviews merge requests and pull requests on GitLab, Bitbucket and Azure DevOps
The same review Xovris gives a GitHub pull request now runs on GitLab.com and self-managed GitLab merge requests, Bitbucket Cloud and Bitbucket Data Center pull requests and Azure DevOps pull requests: every finding lands as a comment on the exact line it is about, a fix that is proven arrives in that host’s own suggestion form so it is one press to accept, a critical finding blocks the merge where the host allows it, and the pre-merge checks report as the host’s own status so a branch rule can require them.
Failing CircleCI and Jenkins builds show up in the review, on the line they point at
Connect CircleCI or Jenkins and the review reads the build of the exact commit it is reviewing: every build is named with its result in the summary, a failure whose log points at a line of the change becomes a finding on that line, and the failing log goes to the fix so it is proven against the same tests – with any password or token in the log hidden first.
Xovris closes the risky doors your router leaves open to the internet, and presses each router setting for you
When the router check finds remote desktop, Windows file sharing, Telnet, FTP or another door that should never face the internet, Close it for me asks Xovris to secure your router: on an OpenWrt router it signs in with the password you give it, shows every change before your yes, rehearses it on a copy, switches the opening off, checks it and can put it all back in one press. In a shared tab of your router’s own page, Xovris now presses each setting after your yes – you only type the router’s password, which never leaves your screen.
Ask Xovris to restart your router – it does it, waits for it and checks nothing changed
You can now say 'restart my router' and Xovris finds your router through the Xovris app on your computer, signs in with the password you give it (never a default or a guess), asks for your yes, and restarts it through the router's own settings. It then waits for the router to answer again and checks it: a fresh start, and every setting exactly as it was - or it tells you which one differs. On a router of any other make Xovris restarts it through its own interface, and asks for one press of its power button only where nothing else reaches it.
Xovris chases a vendor for you, and vendors can send their fixes straight in
When a fix depends on a vendor, Xovris follows it up with them and keeps you told where it stands. Vendors can also publish their own fixes into Xovris; each one is checked before it is ever offered to you, and credited to the vendor.
Software makers can send their own fixes to Xovris, and every customer they help gets them once proven
A company that makes software or devices can now send Xovris its own fix – an upgrade, a configuration change or guidance – straight from its Xovris account or its release pipeline. Xovris takes public knowledge only, never anything about a customer, holds each fix once however many companies send it, and rehearses it in fresh copies before anyone is offered it. The sender sees each fix waiting, being rehearsed, proven or refused with the reason, and can withdraw one in a press.
A code fix reaches you only once your own tests pass with it
When the fix is a change to your code, Xovris first shows the failure with your repository’s own tests, and offers you a change only once those tests pass with it – the failing test before and the passing test after, side by side.
A fix is called fixed only when your real system passes afterwards
Xovris closes a fix only on a test of your real system after the change – never on a stand-in for it – and every control you can press in Xovris first checks that what it controls is really there and answering.
Xovris reviews a code change before it merges – where you work or on the pull request
You can now run npx xovris review and Xovris reviews the changes you have staged, not yet staged or already committed, right where you work. On a pull request it checks the title and the description against what the change really does before anyone merges it, and it follows your repository’s own guideline files, so the review holds the change to the rules your team already wrote down.
Xovris can switch off the feature flag behind a fault – in LaunchDarkly, ConfigCat and Statsig
When a fault started with a feature flag, Xovris turns that flag off itself, within your policy, then reads the flag back to show it really changed – and one press puts it back. It works with LaunchDarkly and now with ConfigCat and Statsig too.
Your problems are only ever worked on by Xovris itself
Only Xovris’s own production systems ever work on your problems: when its AI investigates or fixes something for you, it runs there, under terms that forbid training on what it sees. Nothing you send ever passes through anything Xovris uses for testing – and if a testing credential ever appeared where your work runs, the AI would switch itself off and say why.
A safety net on the device before any change to it
Before Xovris changes a computer, server or network device, it takes that system’s own safety net first – a snapshot or restore point, or a saved configuration on network gear – and checks it. If the safety net cannot be confirmed, the change waits for you instead of running on its own.
Every fix comes with a signed receipt you can check yourself
Each fix Xovris makes now comes with a receipt signed by Xovris’s key. Download it from the fix and check it with a standard tool – no account and no Xovris needed. Change a single character in it and the check fails, so the record you hand your auditor or insurer is exactly the one Xovris wrote.
Every fix Xovris proves comes with a signed proof you can check yourself
Each fix Xovris proves now carries a signed proof for your records: the cause it proved, the failure shown before the change, the change itself, the test after it and the independent check’s verdict – with what it proves and anything still owed said plainly. You, your auditor or your insurer can download it from the incident and check the signature without asking Xovris.
See how often Xovris fixes things like yours, from your own record
Your account now shows, for each of your systems, how often Xovris connected, diagnosed, fixed first time, verified and rolled back – counted from your own fixes only, with a rate shown once there are 30 to go on and the last failure named. Ask whether Xovris covers a list of systems and the answer comes from your own record too.
A fix that failed on your systems is never tried there again
If a fix has failed or been undone in your environment, Xovris keeps it out of the next attempt there and records it as what not to do – whatever happened with it anywhere else.
A person reviews a sample of Xovris’s results before any of them counts
Before any result counts towards a 100% figure, a person checks a sample of it. A result still waiting for that check, or one a person rejected, never counts.
Take back access and a change under way stops before its next step
Every request and every automatic step carries who it acts for, their role, the autonomy you allowed, the budget, your consent, your data region and any stop switch – worked out by Xovris itself, never taken from the request. It is checked again before each step of a change, so removing a person, lowering autonomy or pressing stop takes effect before the next step runs.
No change runs on an out-of-date view of your system, and one that cannot be undone always asks you
Before Xovris writes anything it reads what is there and compares it; if it changed in the meantime, the plan is made again. A change that cannot be undone, or only by hand, is never made under a standing rule – a person approves it every time. Every action says what it may cost and how it is reversed, and AI tools that use Xovris see which actions look and which act.
Every change says what it touches and how it is undone – before it runs
Each change Xovris can make now carries its exact target, the checks that must hold first, how long it may take and how it is put back. A change with no automatic way back never runs without you, and every change leaves a receipt of what was there before and after.
Every fix on autopilot, on every plan that includes fixing – proven before it runs, and put back if it does not hold
Xovris now fixes on its own on every plan that includes fixing. A fix already proven by its past runs straight away. One that is not is either tried first on a copy of your setup and runs only once that passes, or runs with a snapshot kept so it can be put back. After every change Xovris checks that it worked and puts it back automatically if not. A change that cannot be undone runs only after it passed on a copy, and a change that pays someone still waits for your one tap. If a fix ever fails for anyone, it is proved again every time it is needed. Each change, and Settings for each kind of change, says how the fix was proven. Only your own settings – Off, Review first, busy periods, a second approver – or a problem Xovris names as a genuine blocker stop it.
Every repair carries its proof, ready to hand to an auditor
Each repair in an incident shows its before and after tests, the independent verdict and a receipt, with its place in your workspace’s tamper-evident record – which you can check yourself at any time.
Nothing is marked fixed unless the newest check passed
An issue counts as fixed only while the most recent check after the fix passes. If a later re-check fails, it is no longer shown as fixed and the work carries on.
The proof of a repair can never be quietly changed
Every verdict, test and receipt is written once and kept in your workspace’s tamper-evident record. A check compares each one with its recorded copy and names anything changed or removed.
Xovris gets quicker at fixing each kind of problem, week by week
For every kind of problem, Xovris tracks how long a proven fix takes and how often the first fix holds, week by week. A fault it has fixed before is fixed straight away with the fix it already proved, and anything that gets slower or worse raises an alarm for Xovris’s own team.
An imported runbook never runs until you approve it
Bring your own runbooks into Xovris and each one stays a draft until someone approves it – nothing runs from an unapproved import. Every vendor page Xovris reads is kept with its address and the date it was read.
See which fixes run on autopilot, and know each change works before it runs
Automation now lists every kind of fix – DNS, certificates, configuration, code, cloud and more – with whether it runs on its own for you and how each change is proven first: by your own past fixes of that kind, every one verified in a row; just now on a copy of your setup; or with your setup captured and the way back rehearsed, so Xovris puts it back on its own if the check afterwards fails. One failure and that kind is proven afresh every time until it earns its record back. You change the choice right there, and before anything is saved Xovris tells you exactly what it will mean. Your choice is never made for you.
Ask for changes to a prepared fix in your own words
Before you approve a fix, say what you want different – “also check the checkout page”, “not before 10pm”, “watch it for 15 minutes” – and Xovris prepares it again with the difference shown. Anything it cannot change is said back with what it can.
Choose, for each kind of change and each thing you own, whether Xovris fixes it on its own
In Settings, set code, configuration, DNS, certificates, network devices, computers, cloud and accounts to “On its own”, “Review first” or “Off” – and override it for any single domain or device. Each choice says exactly what it lets Xovris do, and every fix is still tested before and after with the way back kept.
No risky change in the middle of your launch: Xovris reads your calendar
Tell Xovris when you are busy – a launch, a sale, Black Friday, Christmas – or connect your calendar, and it holds every risky change in those hours unless you allow that kind or approve the exact fix. It reads only the events whose words say they matter, so a team lunch never holds anything, and if it ever cannot read your calendar it keeps the freezes it already knows and tells you why.
One bar on every open incident: fix it, show me first, call me or handle it
Every open incident ends with the four choices a person actually makes – let Xovris fix it, see the fix first, be called, or have Xovris handle it from now on – the same from the app, the alert email or Slack. Each answer says exactly what happens next and why, and a fix only runs on the exact version you were shown.
A DNS fix changes only the record it says, and your email logo's certificate is watched
When Xovris puts a broken website record back, it proves afterwards that nothing else in your domain moved – every mail, verification and security record exactly as it was – and tells you separately when your DNS company accepted the change and when the internet caught up. If your domain shows a logo beside its email, Xovris now watches that logo's certificate and tells you weeks before it expires, or straight away if it is longer than the limit mailbox providers accept.
Every problem goes straight to the repair that fixes it
The moment Xovris finds what went wrong, the incident goes to the repair that fixes that kind of problem – a moved DNS record put back, a crashed server redeployed, an expired certificate renewed, a provider outage noticed and its support ticket drafted – and runs until the fix is proven. When a step needs you, it arrives prepared: one thing to connect, sign or approve.
Every problem ends fixed, with the report you keep – or with the one thing in the way
When something breaks, Xovris keeps going until the check that failed passes again, putting back anything that did not help before trying the next way. You get a plain report of what broke, what changed and the proof it works – or, if only the law, safety, your yes or another company can unblock it, exactly that and what would clear it.
A Fix Report in plain words for every fix
What was wrong, how it was found, what changed and the before and after side by side – plain words first, the technical line under each, built only from the incident’s own records.
Every incident has a live tracker, like a parcel delivery
Detected, cause found, fix ready, approved, applied, verified – each step appears when its record exists, with its time, so you always know where a problem stands.
Xovris keeps going until it is fixed – and shows every attempt
When a fix is refused or does not pass its re-check, it is put back and Xovris tries the next way – as many as it takes. Every attempt appears on the incident as it happens, and it stops only where your safety settings, your consent or the law say so – and tells you which.
Each kind of problem gets the right kind of repair
Xovris sends each problem to the repair that fits it – a hosting outage through your host’s own recovery, a code fault through a change to your code – and in your account you see, for every problem it fixed, which kind of repair it used and how it was proven.
Every fix report says whether Xovris had met the problem before, and the time that saved
The report you keep for each fix – by email, in Slack and on the incident – now shows the past cases like it and why each is like it, never with another customer’s details. Once the fix is proven it shows how long it took against first-time fixes of the same kind here, and when either time is missing it says which, never an estimate.
When a fix needs you to pay someone, you see exactly what and pay them directly
If a fix needs a payment only you can make – renewing a domain at its registrar, a paid certificate, a provider’s higher tier – the alert and the incident show the amount in that company’s own currency, who it goes to, why, and the time to pay by in your own time zone, with one press to their own page. Xovris never takes the payment or keeps your card, and finishes the fix straight after.
A fix counts only once your own problem is gone
Your own failing outcome is reproduced before anything changes and run again after the fix, with the checks around it, and re-checked later; a fix that does not pass is put back and the next approach starts, and you pay nothing for it until the after-test passes.
Fixes that run on their own
A kind of fix may only run without asking you once it has been proven to work, every time. Until then – and for every new version of a fix – Xovris prepares the whole change and runs it as your policy sets - on its own where the policy allows, held for you where it keeps the step - and the change tells you why.
A problem still unsolved after a day stays one Xovris keeps working on
Every problem still unsolved a day after it was opened stays open and visibly in progress until Xovris proves it solved – and you pay nothing for the fix until then. Your account shows each one and how long it has been open.
Busy periods, a second approver and earned trust for fixes
Tell Xovris your busy days – Black Friday, Christmas and New Year in one press, your own dates, or a calendar it reads for you – and risky changes wait for you then. On the Business plans, ask for a second person’s approval on the kinds of change you choose. Accept or decline Xovris’s suggestions to do more on its own, each shown with the record that earned it.
Ask “why?” about any fix and see what Xovris used
Every attempt in a fix report opens what Xovris used for it and whether each part still holds, beside past cases like it and the check that decided it. Memory lists every code fix Xovris tried in your repositories, the change kept and the clean re-run that proved it.
Password-only sign-ins found, and an affected account locked down in one step
People and access lists every account that signs in with a password alone, by person and system, each with its fix. An admin’s account is raised in severity. When an account is affected, one button ends every session and asks for a new password in each system that holds it. A dry run first names who and where, and your policy allows or refuses each step. If an address confirmed as yours turns up in a breach, you get one alert saying what happens next.
Xovris suggests letting a kind of change run by itself once your own record earns it
When a kind of change you keep in review has a clean record on one thing – by default 10 changes applied and none rolled back – Xovris suggests making it automatic for that thing, never for a kind you switched off. Accept or dismiss it; accepting sets just that kind on that thing.
Emergency mode: one button when something is badly wrong
Hacked, held to ransom, defaced or locked out? One button opens one critical incident, named in plain words for what happened. Xovris keeps what it holds at that moment as evidence a later change cannot rewrite: the DNS records it last verified, the week’s changes and the sites it watches. It says the steps to contain it, which restores Xovris can and cannot run, and who was told.
A fix counts only when your own journey works again, and one that does not is put back
After a repair your own journey is run again – the site’s address with the status and words it must answer, or the check that first failed – and only that pass marks the incident fixed and verified. A repair that fails its after-test is put back to the state captured before it, the earlier state is confirmed, and the incident stays open as rolled back. Fix applied and service restored are always two different states, and a wrong cause or a broken fix is never passed as done.
Xovris puts known configuration faults right itself, and says the exact value it writes
Xovris puts known configuration faults right itself. A DNS record goes back to the value it held when it last answered healthy. Missing security headers are set at the edge, only the missing ones. DNSSEC is switched on. Each change is captured first, applied, checked and can be undone exactly, and Xovris says what it writes and the value before it writes it. For a domain whose DNS is connected, Xovris checks SPF the way receiving mail servers read it. It writes the exact record for each finding: one where the domain sends mail and has none, ~all in place of +all or ?all, and one record in place of two or more.
A fault that keeps coming back gets a lasting fix proposed
When the same kind of fault has been fixed three or more times in 90 days, Overview suggests a way to prevent it, saying how often it was fixed and since when, beside your other risks. It is offered once, accepted with one tap and settled when done.
Routine computer faults repaired within the access you gave
On a computer running the Xovris app, a routine fault is found and fixed within the access you gave – the firewall switched back on for exactly the profiles that were off, and the finding gone on the next check. The same fix on a healthy computer leaves it exactly as it was, and a fix whose after-test fails is put back as it was.
Every read, change and security action Xovris makes is on a record nobody can edit, with its way back
Each step is one entry saying why Xovris took it, the state it changed from and how to take it back. A security action – an attacking address blocked, say – is recorded when it is applied and again when its after-test proves it, and a block whose after-test fails is recorded too. No entry can be changed or deleted, and if anything alters the record behind the database’s own guard, Xovris names the first entry that no longer adds up.
A wide change goes to one place first
A change that touches several places goes to one first, is checked against your own uptime and error rates, and is put back automatically if anything gets worse – the other places are never touched.
Every change is weighed once, before it runs
Each change is judged once by how far it could reach and whether it can be undone, then goes ahead, asks you once, needs two people or waits – never an endless chain of approvals.
Repairs run one recorded step at a time, and anything that does not hold is put back
A fix runs only within what you allowed, each step recorded; if it does not pass its re-check, the change is put back exactly as it was and the next approach starts – the incident stays open and visibly progressing until a fix is proven.
Fail over to your standby while a provider is down
Name a standby for the record your site follows – a second origin or provider – and the next time the provider in front of your site has a confirmed outage, Xovris points your site at the standby once it answers healthy, checks your site, and points it back when the outage is over. Every change Xovris makes – a DNS record added or removed, a certificate installed – keeps the way back it proved before it started.
A fix is proved on the request that failed – and on the pages around it
After you change something to fix an error, Xovris re-runs the exact request that failed and each workflow next to it, and keeps watching through a window before it calls the issue fixed. A fix that breaks something nearby is caught before it is called done.
Code fixes that fit your codebase
When Xovris fixes an error in your code, the change follows your team’s own conventions, and anything your team rules out is never tried. Every attempt is kept beside the pull request, with the test run that proved the one you see.
The other areas
Words on this page, in plain English
Each technical word this page uses, said simply.
- Domain Name System (DNS)
- The internet’s address book: it turns a name you type into the number of the computer that answers for it.
- Program-to-program connection (API, APIs)
- A way for one program to ask another for information or to do something, without a person clicking through a website.
- Container system (Kubernetes, K8s)
- Open-source software that runs an app’s parts across many servers and restarts them when they stop.
- Written steps (runbook)
- The steps to follow for a known problem, written down so anyone on duty can follow them.
- Signed address-book answers (DNSSEC)
- A signature on a domain’s address-book answers, so a computer can tell they were not forged on the way.
- Allowed senders record (SPF)
- An address-book record listing the servers allowed to send email as the domain.