Investigation
Finds the cause on its own.
Xovris checks each likely cause against real readings, rules it in or out, and names the one that explains what you see.
XovrisOn its own
- A lit square is a feature people use today
- Newest: Each problem shown failing, explained and fixed
Every feature in Investigation, 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
Each problem shown failing, explained and fixed
Every problem page draws the check failing, the cause and the fix as one sequence.
The problems library, redesigned
The problems index and every problem's page read clearly on a phone and a desktop, and nothing moves under reduced motion.
Find your problem type at a glance
The problems index opens on its reference in the first screen, the kind with the most problems first, so you find yours without scrolling.
See each step Xovris took to find a cause
Every answer carries its trail - each check Xovris ran, what it read and what it found - folded away until you open it.
One request's failures become one incident
When one request fails across your apps, servers, databases and security tools, it opens as one incident, not four.
Failing beside healthy, the healthy window named
A failing period is shown beside a healthy one, and Xovris names which window it compared against.
See what changed just before a failure
On an incident, see what changed just before it - a DNS change first for a name that stopped resolving, a deploy first for an app that broke.
What changed before every failure
Every failure carries the changes that came just before it, the likeliest cause first.
One timeline from every source
For an incident, Xovris rebuilds one timeline of what changed from every source that records change.
Timelines in true order across clocks
Events from many sources are put in the order they really happened, even when their clocks disagree.
Incidents investigate themselves
When something qualifies as an incident, Xovris starts investigating it on its own.
Investigations start within 30 seconds
Every incident's investigation starts within 30 seconds of the event that raised it, even in a storm of incidents.
A critical incident never waits in a storm
Even when other workspaces are flooded with incidents, a critical one in yours starts at once.
Room kept for critical incidents on every plan
Every plan keeps room for a critical investigation, so one never waits.
Website failures named for what they are
When a website fails, Xovris names which kind of failure it is - nine kinds told apart - in the answer you read.
Problem pages read in plain words
Every problem page reads in finished sentences - no template text shows through where a detail goes.
Slowness traced to its real cause
Xovris tells CPU contention, memory pressure, a slow query, a lock wait, a slow outside service and a bad deploy apart, then tunes, rolls back or scales to fix it.
Healthy and slow requests compared side by side
Xovris compares healthy and unhealthy requests and names the attribute that explains a regression.
The attribute behind a slowdown, named with a significance test
Slow and healthy requests are compared automatically with a significance test, and the attribute that explains the difference is named.
A slowdown hitting a few users explained
A p99 slowdown that only affects a small group is found and explained from kept traces - the group, the cause, and what sampling left unseen.
Visitor feedback linked to their session
A visitor's note is linked to their own session, so Xovris can see what they saw and diagnose it.
Every open incident shows when it will be fixed
Each open incident shows a live 'fixed by' time worked out from your own similar fixes. Nothing is promised from fewer than three.
Any error or log line read in plain English
Signed in, or on the free tool with no account, paste an error and read in plain English what it means, its likely causes, what to do now and where that comes from. An error Xovris does not recognise is said to be unrecognised, and costs you nothing.
A broken build fixed and proven before CI runs again, and every change checked for attack paths
When a build breaks on GitHub Actions, GitLab CI, Azure Pipelines, Bitbucket Pipelines, CircleCI or Jenkins, Xovris reads the log, says what went wrong in plain words, and where a fix is known it makes the fix and runs the job's own command in its sandbox first - where no fix is proved safe yet, Xovris says what it found and the next step it is taking. Every change is also checked for the attacks that matter: database queries built from input, the server fetching an address a caller chose, records read by id with no owner check, outside text written into an AI's instructions, and a check and an act that are not atomic - each held to whether a request can reach it, look-alikes set aside with their reason, and a merge blocked at your threshold.
Every incident, fix and report is one card you can trust, everywhere you read it
Each incident, fix, issue and report is now one card with its verdict, its proof and how sure Xovris is – the same card in the app, from the API and in your own AI tools, live from real state and complete without motion. Changes that are not code – DNS records, email authentication, headers, firewall and configuration – are shown before and after, like a code diff.
Every error, slow, rare and new-release trace is kept – and opens on its log line
Xovris keeps every trace that matters – each error, each slow or rare request and the first ones after a new release – and opens it in place on its log line as a waterfall of what really happened.
Code review that knows what your change reaches – files, services and the live site
A review now reads a large change in the order it builds up and names everything it reaches: the files that use it, the services they belong to and the live domain or host they serve. It checks a change to your DNS against the zone as it is live and names the mail or site outage it would cause, flags an API your declared version deprecated or removed with the publisher's dated page, names every caller a contract change breaks in your other repositories, merges every analyser that speaks SARIF into one review with no duplicates and no secret shown, answers a question about the change with the exact lines, plans an issue and judges the finished pull request against that plan, turns a finding into an issue in Jira, Linear, GitHub or GitLab, checks a pull request against its issue's acceptance criteria, and takes its review rules from a file in your repository, each mistake named by its line.
Every incident shows what changed just before it
Each incident page now carries “What changed just before” – the deploys, merges, commits and DNS changes that came before the error – so the likely cause is in front of you before you go looking.
Every answer, fix and report shows what Xovris found, in one shape
Whatever Xovris does for you – an answer, a fix, a code change or a report – now arrives in the same shape: what it found, the evidence it rests on, how sure it is, what changed and how to undo it. It reads the same way on your phone and at your desk, in light or dark.
See how fast Xovris found the cause and proved the fix
Every issue now says how long Xovris took – for example “Cause found in 42 seconds” and “Fixed and proven in 6 minutes” – and Overview shows, for each kind of problem, the usual time and the slowest one in ten, so you can hold Xovris to its speed.
A name of yours pointed somewhere new is checked as a possible hijack
If one of your names starts pointing to a place it has never served from, and nobody confirmed the change, Xovris raises it as a possible hijack with what to do next: put the record back and change your DNS provider’s keys. A new destination you confirm as yours is remembered for you, so a change you made on purpose is not raised again.
Facts first: a provider outage that cannot touch your site is not blamed for it
Xovris settles what it can from facts before anything else. When a provider you use reports a problem, Xovris checks whether that problem touches the path your own requests take; one that does not is set aside, so you are not told it is someone else’s outage when the cause is in your own setup – and the fix it proposes is still tested before and after.
What Xovris takes to work on a problem is deleted the moment the job ends
When an incident closes, whatever way it ends, everything Xovris gathered to work on it – the readings it took and the answer to each step – is deleted at that moment. The incident keeps a note that it existed and says when it was deleted. What Xovris remembers about your setup stays yours to see, export and delete.
On a problem it has never seen, Xovris names a cause only when your own systems back it
When something goes wrong that Xovris has not met before, you get the cause your own systems support – never a guess. When the evidence still fits more than a single cause, you are told each, plainly, rather than one picked for you – and Xovris keeps working within what you allowed until the fix passes its check afterwards.
Xovris follows the failing request back to the change that broke it
When something breaks, Xovris starts from the request that failed and walks back through what changed around it – a record, a deploy, a setting – and names the cause only when it can show how that change led to the failure. If it cannot show it, it says so plainly instead of guessing, and either way you see the evidence it used.
See everything Xovris knows about your systems, and why it acted
The Memory page shows every fact Xovris keeps about your systems – what kind of fact it is, when it became true and when it was recorded, how sure Xovris is, whether it may be out of date and why, and the decisions it influenced, each linked. Correct a fact, forget it with a receipt, or export everything; a fact of a kind the page does not name yet is still shown.
Xovris names the company behind a failure, even one you never connected
When your site fails because of a company it depends on – your DNS at one provider, your hosting at another – Xovris names that company and checks its own status for you, instead of stopping at the one you connected. When one problem is causing another, you are told about one problem with its consequence, not two separate ones. And when your checkout breaks while your home page is fine, you are told it is your own pages, not your host.
A lookalike page or a forged link never becomes a fact about your incident
An outage notice is taken only from the provider’s own status page. A page on a lookalike address, or a link made to look like one, is refused and named in your incident’s record, so nobody can steer Xovris with a fake notice. And a brief blip where something could not be reached never decides the cause on its own – Xovris looks again before it tells you anything.
When only your computer cannot reach a site, Xovris says so and starts there
If the Xovris app on your computer cannot reach a site while Xovris’s own checks reach it fine, one problem is opened from your side, naming your computer, and the look into it starts with your own connection – so you are not told the site is down when it is not.
The record of what Xovris’s AI did never contains your words
Every record Xovris keeps of its AI’s work follows OpenTelemetry’s open standard at a fixed version, and a record that would carry your question, a message or a reply is refused whole rather than trimmed – so the record of what happened never becomes a copy of what you said.
Answers point to the maker’s own documentation
When Xovris explains a fault in something you use – Cloudflare DNS, Let’s Encrypt certificates, Google and Microsoft mail settings, PostgreSQL, Kubernetes and more – it links the maker’s own documentation page it relied on, so you can read the source yourself.
An incident closes only when its cause is proven
If service is back but the cause is not yet proven, the incident stays open as service restored and closes on its own once the cause is proven. "Unknown" is never accepted as a cause.
Correct Xovris once and the correction sticks
When you correct Xovris, or an approach does not work, it is checked again every six hours against what really happened, and the same failing approach in the same place is caught and stopped.
Explain any incident to your manager, a client or your family in one press
Every incident now has an Explain it to someone button. Choose who it is for and Xovris writes what happened, how serious it was, what it did and whether the fix was proven, in plain words from the incident’s own record – ready to copy, share or download.
Nothing of yours leaves your workspace
When a problem like yours has a fix that has already been proven, Xovris brings it to your incident first – never with anyone’s names, data or settings, and anything that stops working is withdrawn on its own. What Xovris works out from your own incidents stays a suggestion until you or an independent check confirms it, and you can see it, correct it or delete it at any time.
A hard problem gets more of Xovris on it – never more than its budget
When a problem does not give way at the first look, Xovris brings in a specialist for every kind of cause still open and works them all at once. Every step is priced before it runs and charged to that issue’s own budget; a step that would go over is held back, and you can see that it was.
An incident says at the top when a change was put back, and whether it has happened before
When a change does not pass its after-test and Xovris puts it back, the incident says so at the top: what failed, how and when it was put back, and what Xovris is trying next – and the issue stays open until a fix passes. Beside the title, every incident now says whether it has been seen here before or is the first time.
See what an incident rests on, and how sure Xovris is of each link
An incident’s page now shows what its address depends on – where it points, who hosts its DNS and its mail, who issues its certificates – with when each link was last seen and how sure Xovris still is of it. A link not seen recently is named first, and the page says when it lowers the diagnosis and what refreshes it.
Your incident workspace follows you to every device
An incident’s workspace is saved as you go and opens on your phone or another computer showing when it was last saved and what changed since; if two devices disagree, nothing is overwritten. Every incident can be replayed from its timeline, and every procedure keeps its versions, each run recording exactly which version it used.
Words hidden in a log, a web page or a ticket never steer Xovris
Anything Xovris reads while it works – logs, error messages, web pages, support tickets, code and the answers other services send back – is treated as evidence about your problem, never as instructions. A line planted to make it change something it should not is read, reported and ignored.
Investigations show you what they looked at, and never go round in circles
Each investigation shows you, in plain words, what it looked at and what it found. If the same step would run again with nothing new found, Xovris tries another way instead. A paused investigation picks up within a minute of the answer it was waiting for.
Export everything Xovris remembers, or keep any fact out of automatic changes
On the Memory page you can correct what Xovris remembers about your systems, make it forget something, export everything it remembers, and mark anything as never to be used for an automatic change.
See which of Xovris’s decisions each thing it remembers influenced
Every fact on the Memory page shows how many of Xovris’s decisions it influenced and links the latest three, so you can see exactly where something it remembers about your systems was used before you correct it or make it forget.
A problem Xovris has never met before still gets worked out
Xovris keeps a list of every kind of problem it can find in each feature and each integration, each with the check that finds it. When a fault is new to it, Xovris works the fix out from the evidence instead of stopping because it has not seen it before.
A change that is hard to undo waits until it is known to be safe
Before Xovris makes a change that cannot easily be undone, or that touches a lot at once, the plan and the evidence behind it are checked again. If anything does not agree, Xovris holds the change and asks you, showing what it found.
Ask what Xovris knew at any moment, and what was true then
For everything Xovris knows about your systems it keeps both when it was true and when Xovris learned it. So after any change you can ask what Xovris believed at, say, 14:02 – and what was really the case at that moment.
Your critical incident is never slowed down by someone else’s
Xovris keeps capacity in reserve for critical incidents – for every workspace and across all of them – so a storm of incidents elsewhere cannot hold up yours, and alerts go out straight away without ever waiting behind an investigation.
Decide where Xovris works on your data, and that nothing is kept
In your workspace settings you choose the region where Xovris processes your data, that nothing you share is kept once an answer is written, and which of our processors may handle it. Xovris checks your choices before every step it takes and never works outside them. When a choice means a step cannot run, it tells you why and what you can change. A release that would ignore any of these choices is stopped before it ships.
When the AI cannot run, the incident says what carries on and when it tries again
If the AI is unavailable, your allowance is used or your own limit is reached, the incident now says so at the top: what keeps running without it (your checks, alerts and the diagnostic tests), what happens next, the time Xovris tries again, and the one thing you can do – opening the exact place in Settings to do it. The problem stays open and keeps moving until it is fixed and proved.
Xovris gets better at each kind of problem, and works on any that slip
Xovris keeps track, for each kind of problem, of how well it fixes and proves it first time; a kind that gets worse is worked on until it recovers, never hidden in an average. You see how it has done on your own systems in your account.
When the cause is out of sight, Xovris says exactly what to connect
If the cause of a problem cannot be settled from what Xovris can see, the incident names each place it was blind and the one connection that would show it – your code host, your cloud audit log, your deploys, your logs – so the next time it can.
Every action Xovris took, with its reason and its way back – on a record nobody can edit
Each incident now lists everything Xovris read, changed, decided or put back for it – why it did it, and the snapshot it can be undone from – and Logs tells the whole life of a certificate, a domain or a server the same way. The record is checked unbroken every time you open it, Xovris’s own staff cannot edit it, and you can save an asset’s story as a file anyone can verify.
Your account shows how Xovris has done on your own problems
For your own systems, see each problem Xovris fixed, whether it was fixed first time and how it was proven – and if something took more than one try, that is shown too, never hidden.
Every incident is investigated before you are asked anything
When something breaks, Xovris works out what happened from your own records and the providers you depend on, and shows you the cause in plain words, with the records that prove it.
A fault Xovris already knows is recognised on its first signal
When a fault is one Xovris already knows how to fix, it recognises it from its first sign and says so – and no customer’s details ever reach another.
A page that loads but breaks for visitors is caught
When a page answers but its own files or scripts fail, Xovris loads it the way a visitor would – on a phone too – and names each part that failed, so a broken script is never mistaken for an outage.
Investigations see across your whole estate
While it works out what broke, Xovris can look across logs, releases, DNS, cloud settings, devices and sign-ins – then makes the fix once the cause is proved.
What Xovris works out about your setup stays a suggestion until it is confirmed
When Xovris works out something new about your setup while investigating – which server is your main database, say – it shows it as a suggestion until you confirm it or your own systems prove it, and only then lets it shape a fix. You can ask what Xovris believed about any part of your setup at any moment, where each fact came from, and what was actually true then.
Read every step an investigation took, and play any fix back in order
Activity lists, in order, every check an investigation ran and every action Xovris took. On an incident you can play each fix back in order: what Xovris looked at and where it came from, what it changed with the before and after, and every test that proved it.
Every diagnosis shows what it rests on, and secrets are masked before the AI reads anything
Each piece of evidence an investigation gathers is kept with where it came from, when it was seen and the step that produced it, and the conclusion names the check that decided it; an answer about your logs cites its lines. Before any of it reaches Xovris’s AI, secrets and personal details are masked, so the masked text is all that is ever stored or replayed.
Every new incident starts its own investigation, and a critical one never waits
When an incident opens, Xovris starts investigating it by itself, once – a repeat of the same alert starts nothing new, and a start that fails is tried again. When many incidents are open at the same moment, a critical incident in your workspace still starts at once on capacity kept back for it, instead of waiting its turn.
Every incident says how bad it is, how sure Xovris is, and which area it sits in
Severity comes from the evidence and how important the affected thing is – a security or data exposure is critical, and evidence Xovris could not read is never taken as no impact; you can lower it with a reason. Every incident also sits in one of eight areas – code, site, DNS and certificates, email, cloud, devices, network or identity – by the same rule that names its kind.
See which changes a coding assistant wrote and how they held up in production
Xovris recognises changes written by AI coding assistants from what each leaves in the change itself, and shows for each one how many of its releases led to a problem and how quickly they were put right – beside the changes people wrote.
Watch logs arrive live, with the line that proves the cause flagged and the stream paused on it
During an incident, every line that matches your filter arrives within about half a second – none of them sampled away. Lines you have never seen before are marked, lines that support or contradict what Xovris thinks is wrong are marked with the reason, and the stream stops on the first line that proves it, so you see the proof instead of scrolling past it.
Write, check, save and share queries over your logs, metrics and traces
Every answer to a question about your logs now shows the exact query it ran, which you can edit and run again. Write your own in SQL or as a short pipeline – logs, then where, then count by service – and Xovris explains it in words and points out the usual mistakes before it misleads you: a time with no time zone, a misspelt service name, a filter that reads everything. Save a query for your team or keep it to yourself; every change to it is kept.
Investigations read the providers your site depends on
When your provider is the cause, Xovris names it and links to where the provider announced it, with the date it was published. An outage at your provider is named as the cause, and your own settings are kept exactly as they are.
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.