EVERY BUSINESS EVENTUALLY BECOMES TOO COMPLEX FOR ITS FOUNDER TO WATCH PERSONALLY.SENTINEL EXISTS FOR THAT MOMENT — AND EVERYTHING BEFORE IT.
Sentinel is autonomous enterprise intelligence and protection: continuous visibility, governed response, preserved security evidence and business continuity. It is introduced early, so the company can grow without losing sight of what it is becoming.
YOU CANNOT PROTECT WHAT YOU CANNOT SEE.AND YOU CANNOT RESPOND TO WHAT YOU CANNOT VERIFY.
Complexity grows. Visibility does not automatically grow with it. Eventually, leadership is responsible for systems it cannot personally observe.
An illustration of the principle: what is not in sight cannot be protected.
The argument in fullThe first statement, the visibility gap
The first statement
YOU BUILT THE COMPANY.SENTINEL WATCHES WHAT COULD TAKE IT AWAY.
A business can lose control long before leadership knows something is wrong.
Illustrative demonstration
A privileged account changes.
A system stops reporting.
Sensitive information begins moving.
A service behaves differently.
A vendor connection fails.
An automation takes an unexpected action.
Each event, on its own
The relationship between them
Six events that each look small, and the relationship that joins them. A drawing to explain an idea: it contains no data and no measurement.
Individually, each event may look small.
Together, they may describe the beginning of something much larger.
Sentinel exists to see the relationship.
EVERY BUSINESS EVENTUALLY BECOMES TOO COMPLEX FOR ITS FOUNDER TO WATCH PERSONALLY.SENTINEL EXISTS FOR THAT MOMENT — AND EVERYTHING BEFORE IT.
The visibility gap
THE VISIBILITY GAP
A company grows by accumulating people, systems and responsibility. A founder’s attention does not accumulate with them.
Complexity grows. Visibility does not automatically grow with it. Eventually, leadership is responsible for systems it cannot personally observe.
Illustrative demonstration
A company grows around its founder
Sentinel layer
The founder
What the founder can see directly
In direct sight
Beyond direct sight
Seen through Sentinel
At scale the company is drawn as a field of points. Each point stands for many: a symbol, not a count.
Early
Founder
Email
Domain
Website
Payments
Cloud
Growth
Employees
Applications
CRM
Finance
HR
Vendors
Infrastructure
Repositories
AI systems
Customer data
Multiple environments
Scale
Hundreds, then thousands, of identities and services
The relationship, drawn. Conceptual, not measured.
Company complexity ↑
Human direct visibility ↓
Visibility with Sentinel
The gapThe gap, closed
Early
Growth
Scale
Step 4 of 4
With Sentinel
A layer that observes what is connected to it, verifies what it knows and brings the picture back to one place. Direct sight does not have to grow for the picture to return.
The founder, email, a domain, a website, payments and a cloud account. Few enough for one person to hold in mind, and to see directly.
Employees, applications, a CRM, finance, HR, vendors, infrastructure, repositories, AI systems, customer data and more than one environment. The founder's own attention has not grown with them.
Hundreds, then thousands, of identities and services. Leadership is still responsible for all of them, and can see directly only a few.
A layer that observes what is connected to it, verifies what it knows and brings the picture back to one place. Direct sight does not have to grow for the picture to return.
A conceptual drawing of how a company grows around its founder. It is not a measurement and contains no data.
GROWTH CREATES DISTANCE BETWEEN LEADERSHIP AND THE SYSTEMS LEADERSHIP IS STILL RESPONSIBLE FOR.
SENTINEL CLOSES THAT GAP.
02
Part 02 of 06
Who it is for
The same need, at every stage of the company.
For founders
YOU BUILT THE COMPANY.SENTINEL WATCHES WHAT COULD TAKE IT AWAY.
The founder is often the most privileged identity in the company. One account can reach most of what the company is. Sentinel is the independent layer watching the systems around it.
One point of view, at the centre of every system the company runs.
For foundersOne account, most of the keys
For founders
FOR FOUNDERS
The founder is often the most privileged identity in the company.
One account. Most of the keys.
Email, domains, cloud, payment infrastructure, repositories, customer information, banking relationships, AI platforms, legal documents, vendor accounts and administrative credentials often meet in one person.
Illustrative demonstration
The founder’s account
Email
Domains
Cloud
Payment infrastructure
Repositories
Customer information
Banking relationships
AI platforms
Legal documents
Vendor accounts
Administrative credentials
One account, drawn with what it can reach. An illustration of concentration: it describes no company and contains no data.
A compromised founder account can therefore represent concentrated organizational risk.
Concentration, not carelessness, is the point. The more one identity can reach, the more one identity’s compromise could reach. Nothing here says it will happen. It says where the weight sits.
THE INDEPENDENT LAYER WATCHING THE SYSTEMS AROUND THE FOUNDER
Sentinel is not one more account for the founder to guard. It is a separate layer with its own record, its own notifications and its own authority model, watching the systems around the founder rather than being one of them. Eight things it does, and how much of each exists in this release.
01 · Privileged activity observation
Privileged actions are among the most consequential events in a company and the hardest to reconstruct afterwards. Sentinel observes them as they happen, so a record exists before anyone has to ask for one.
Built in partWhat exists today
Built for what reports to Sentinel: every privileged action taken through Sentinel is written to the Ledger in the same step as the change. Connecting a company's other platforms, such as email, cloud and banking, is a design direction and not a shipped integration.
02 · Guard authority boundaries
Authority should have edges. Guard holds each person, session and service to the scope it was given, and the strength of what Sentinel actually witnessed sets a ceiling on how hard it may respond.
BuiltWhat exists today
Tenant isolation at the data path, graded operator authority, session revocation, attestation-bounded response and provider-confirmed network containment.
03 · Anomaly correlation
One odd event is rarely the whole story. Sentinel is designed to read related observations together, so a privileged session, an unusual source and a system going quiet are seen as one sequence rather than three separate notices.
Built in partWhat exists today
Built in part: an incident holds the timeline of its related events and its playbook tasks. Recognising several sources going dark in sequence as one pattern is a design direction for Watchtower, which today evaluates each deployment on its own.
04 · Secondary security notifications
A notice sent only to the account that may itself be compromised is a weak safeguard. Security notifications are being built to go to a separately verified security mailbox, such as security@ or soc@, so the people who need to know do not depend on the founder's own inbox.
In buildWhat exists today
In build, and not yet part of this release. Today Sentinel queues notifications for operators and delivers them by email with bounded retries; a failed delivery is shown as failed. Delivery is never described as guaranteed.
05 · Weekly reporting
A founder should not need to read logs to know where the company stands. The weekly report is designed to say what changed, what requires attention, what Sentinel did, what remains open and where to verify it.
In buildWhat exists today
In build, and not yet part of this release. The design sends the summary to the verified security mailbox and keeps the detail, and the means to verify it, in Sentinel Command.
06 · Command visibility
One place to see the state of the estate and of Sentinel itself, so a founder looks at one picture instead of assembling one from six tools.
BuiltWhat exists today
The authenticated console for events, incidents, visibility, evidence, Range, Safe Mode, continuity, Shield and team. The public Command demonstration runs on fabricated data and says so.
07 · Evidence history
For what Sentinel observes, the record already exists when something happens. Evidence keeps what happened, how it happened and what is true now, so the first question after an incident is not where to begin.
BuiltWhat exists today
A per-tenant hash-chained Ledger, Trace, Pulse and the evidence object model. Evidence object storage is verified per deployment; without verified storage a deployment refuses evidence.
08 · Continuity awareness
Protection is not only prevention. Sentinel asks for state to be preserved at the moment of detection, before remediation can overwrite it, and checks that a recovery has been verified before an incident may close.
Built in partWhat exists today
Preservation requests linked to the incident, outcomes recorded to the Ledger, and a recovery-verification gate on closing an incident. There is no live COLOSSEUM integration yet.
YOU SHOULD NOT HAVE TO BE AWAKE FOR YOUR COMPANY TO BE WATCHED.
Illustrative demonstration
You
Sentinel
Awake and at work
Asleep or away
Watched by Sentinel
An ordinary day, drawn. A person is awake for part of it. Watchtower’s scheduled pass runs every minute, whether or not anyone has Command open.
Delegate without becoming blind
Every growing company hands authority to other people and other systems. The question is whether the founder can still see what was handed over. Sentinel turns PERSONAL CONTROL into GOVERNED CONTROL.
Illustrative demonstration
From personal control to governed control
Personal controlGoverned control
IdentityWho is acting
ScopeWhat they may reach
PolicyWhat the rules say
AuthorityWhat they were granted
DecisionAllowed, constrained, challenged, refused
FounderWho actsWhat they reach
A person
A service or automation
What they reach
The founder’s line of sight
Step 4 of 4
Personal control becomes governed control.
At the boundaries Sentinel governs, every actor passes identity, scope, policy and authority before a decision is made, and the decision is recorded. The founder can delegate without becoming blind.
One person holds every credential and makes every decision. Control is personal, and for a while that is enough.
Delegation begins. Each person holds part of the company's authority, and the founder can no longer see every path to every system.
People, contractors, services and automations all hold some authority. Personal control no longer describes what is true.
At the boundaries Sentinel governs, every actor passes identity, scope, policy and authority before a decision is made, and the decision is recorded. The founder can delegate without becoming blind.
A conceptual drawing of authority as a company grows. Fifteen actors stand for hundreds. The path shown, identity, scope, policy, authority and decision, is the one Guard applies at the boundaries Sentinel governs. Nothing here is a measurement.
Institutional memory, before the institution
Young companies keep too much operational history in people’s heads. Sentinel begins creating institutional memory before the business has a large institution around it, so these questions have somewhere to be answered from: the Ledger, the Trace and the Pulse.
Which Sentinel record answers each question a young company otherwise answers from memory
The question
Ledgerwhat happened
Tracehow it happened
Pulsewhat is true now
Why did we block this?
Primary source
Supports
Not the source
Who changed that?
Primary source
Supports
Not the source
Why did production fail?
Supports
Primary source
Supports
What happened before the incident?
Supports
Primary source
Not the source
Who authorized recovery?
Primary source
Not the source
Supports
And what is true right now?
Not the source
Not the source
Primary source
Where the answer lives
What supports it
The answers come from what Sentinel observed and what was done through it. Where evidence storage is not verified for a deployment, Evidence says so and refuses evidence rather than pretend. The record makes a silent edit to one entry detectable; it does not claim to settle questions of legal admissibility, which belong to counsel.
A founder does not need a security department to begin.
EARLY ACCESS TO THE SAME SENTINEL OPERATING PHILOSOPHY.
BUILD THE PROTECTION LAYER BEFORE YOU NEED TO RETROFIT IT.
Small companies often have few specialists and broad founder privileges. Their systems may be smaller. Their failure tolerance can be smaller still.
Protection from the first day, before the estate has grown.
For new companiesBuild the protection layer before you need to retrofit it
For new companies
BUILD THE PROTECTION LAYER BEFORE YOU NEED TO RETROFIT IT.
Small companies often have few specialists, shared responsibility and broad founder privileges. They deploy quickly, their processes are still forming, and there is rarely anyone whose whole job is security.
Their systems may be smaller. Their failure tolerance can be smaller still.
The shape of an early company
Smaller systems. Smaller failure tolerance.
A young company that loses control of its domain, payment account, customer database, repository, identity provider or cloud environment can experience an existential interruption. That is not a prediction. It is the reason the order matters: early architecture reduces future complexity.
How an early company is built
01Few specialists
02Shared responsibility
03Broad founder privileges
04Rapid deployment
05Processes still forming
06Limited security headcount
What a lapse of control can interrupt
DomainCustomers and email stop finding the company.
Payment accountRevenue stops, or moves somewhere it should not.
Customer databaseCustomer information is exposed or gone.
RepositoryThe product and its history are changed or lost.
Identity providerThe front door to every other system changes hands.
Sentinel observes from the day it is introduced. It does not reconstruct the years before. A company that begins early has a record of how it grew; a company that begins late has to find what it already has, and starts its record from there.
Illustrative demonstration
At founding
On the record
Systems join as they arrive
First hires
On the record
Systems join as they arrive
First customers
On the record
Systems join as they arrive
At scale
On the record
Systems join as they arrive
On the record from the first change
Before the layer: history lives in people’s memory
Joined as it arrived
Found and brought in after the fact
At founding. Every system joins the layer as it arrives, and the record starts with the first change. Nothing has to be found later.
First hires. A few systems and accounts already exist. They are brought in once, while there are still few of them. The record starts here; what came before is remembered, not recorded.
First customers. Customer information, a payment account and vendor connections exist before the layer does. Each must be found and brought in, and the history before this point is not on record.
At scale. A large estate has to be mapped after the fact: every system, identity and path found, then brought under the layer. The record starts here, and the years before it live in people’s memory.
Startup economics
SERIOUS PROTECTION SHOULD NOT BEGIN ONLY AFTER A COMPANY BECOMES LARGE.
$99/monthor $990/year
Startup Program: $99/month or $990/year for the first 12 months, on 1 protected site with 3 seats, then Essential.
The Startup Program is a program, not a trial: a program price for eligible early-stage businesses, on the same operating loop the rest of Sentinel runs.
EARLY ACCESS TO THE SAME SENTINEL OPERATING PHILOSOPHY.
Startup ProgramFirst 12 months
Essential$249/month or $2,490/year
Which capabilities each plan includes is stated on the pricing page.
Illustrative demonstration
Startup Program1 site3 seats
The Startup Program: one protected site, three seats, the whole loop.The same loop, wider: Essential, Professional, Enterprise.
1ObserveIs the protected site reporting?
2VerifyIs what arrives what was expected?
3DetectHas something changed, or gone quiet?
4ProtectDoes every action stay within its authority?
5PreserveIs there a record of what happened?
6RecoverCan the company return to a known state?
The security-team gap
The team a young company does not yet have.
A young company may not yet employ a CISO, a SOC, security analysts, incident response, risk analysts, security engineers or forensic analysts. Sentinel does not claim to replace those professions. It automates and unifies part of the observation, verification, correlation, evidence, reporting and governed-response work they would otherwise have to build by hand.
Illustrative demonstration
Professional functions that would otherwise build it by hand
CISO
SOC (carries part of this work)
Security analysts (carries part of this work)
Incident response
Risk analysts
Security engineers
Forensic analysts
Observation. Watching every source, continuously, including the quiet ones. Watchtower holds the expectation of what each registered deployment should report, and notices when a source stops.
Verification. Checking that what arrives is what was expected before anyone acts on it. Sentinel admits and attests what arrives, and keeps the strength of the evidence attached to the decision.
Correlation. Reading separate facts together. Sentinel relates what it sees across systems, so a person starts from one picture rather than from a list of alerts.
Evidence. Preserving what happened, in order, before it can be lost. Ledger keeps what happened, Trace how it happened, and Pulse what is true now.
Reporting. Telling leadership what changed and what needs attention, without a log to read. The weekly report is being built: a summary to a verified security address, the detail and the verification in Command.
Governed response. Acting within a boundary, and being able to show why. Guard decides within the authority its evidence allows and records each decision; anything stronger needs a named person.
That is economic leverage. The groundwork is already in place when the first specialist arrives, so they begin with a working picture instead of a blank page.
Trust
The questions customers, investors and partners ask.
A young company can answer them from a record rather than from memory. Each answer is a Sentinel surface, with its status stated plainly.
Illustrative demonstration
Answered byProtection BannerBeing built
A visible mark on the protected property says that Sentinel is reporting on it, and says so only while that is true. If Sentinel cannot verify reporting, the mark degrades. If the registration is revoked, the claim disappears.
AXIS SentinelMonitoring active
AXIS SentinelNot reporting
Registration revoked: the claim is removed
The mark states what is true: reporting, not reporting, or nothing at all.
Today the visitor seal follows this rule and reads “Monitored by AXIS Sentinel” on a registered page. The Protection Banner is being built. It is visible trust, not a certification.
Watchtower holds an expectation of what every registered deployment should report and compares it with what actually arrives, on a schedule of its own.
Expected
Received
Every expected signal arrives on its rhythm
Expected against received, signal by signal.Watchtower→
Answered byCommandBuilt
Guard decides within the authority its evidence allows and records the decision. Command shows the decision, the containment state and the authority it required, and carries a person’s decision back out under recorded authority.
Five systems, one view
Every system reports into one view, and a person decides from it.Sentinel Command→
Answered byEvidenceBuilt
Ledger keeps what happened, in order. Trace reconstructs how it happened. Pulse shows what is true now, derived from the record. It is a record of what Sentinel observed and did, not a legal finding.
Silence is not assumed to be health. A source that goes quiet becomes a finding, and the disappearance is shown as a gap rather than hidden. Watchtower reports on its own health separately, so the watcher is itself watched.
Expected
Received
A gap: the signal that did not arrive is drawn, not hidden
Loss of visibility is itself a security event.Watchtower→
Answered byCommandBuilt
Operators act under graded roles. Every privileged decision is made on the server and re-validated there, so the console can never grant what the API would refuse, and the authority each action required is recorded.
A named person decides anything stronger
Authority ceiling, set by the evidence
Guard decides and records, within it
Authority has a ceiling, and the ceiling is set by the evidence.Sentinel Guard→
Answered byWeekly reportBeing built
A weekly summary goes to a verified secondary security address, set out as what changed, what requires attention, what Sentinel did, what remains open and where to verify it. The detail and the verification are in Command.
What changed
What requires attention
What Sentinel did
What remains open
Where to verify it
Leadership reads a summary, not a log.
Delivery of any email cannot be guaranteed, which is why the verification lives in Command.
THE PROBLEM IS NO LONGER THE ABSENCE OF DATA.IT IS THE ABSENCE OF CONTEXT.
Every system in the company knows something. Leadership needs what the facts mean together.
Separate tools drawn into one operating layer, from the edge inward.
For established businessesOne operating picture from many systems
For established businesses
ESTABLISHED BUSINESSES DO NOT LACK TOOLS.THEY LACK ONE COHERENT OPERATING PICTURE.
An established company runs an identity provider, email, cloud services, endpoint tools, a firewall, payment systems, a CRM, an ERP, an HRIS, backup, repositories, vendors, monitoring, ticketing and increasingly AI systems.
Each of them is good at its job, and each knows something. What leadership needs is what the facts mean together.
From fragments to one picture
Every system knows its part. No single one is placed to see the whole.
Illustrative demonstration
Sentinel CommandOne operating picture
Step 1 of 3
Each system holds its part
Each of these systems does its job and holds a true part of what is happening. None of them holds the whole.
Step 2 of 3
The facts are read together
Sentinel reads what each system recorded side by side, so a change in one place is set against what the others saw.
Step 3 of 3
One operating picture
Command shows what changed, what it means together and what needs a person: one picture, instead of a separate view per system.
Choose a system to see what it knows, and what it cannot say on its own.
The systems shown are the kinds an established company typically runs. The demonstration is simulated: nothing here reads a live system, and no vendor is faulted. Each does what it was bought to do.
Data is not the gap
Data and context
THE PROBLEM IS NO LONGER THE ABSENCE OF DATA.IT IS THE ABSENCE OF CONTEXT.
Data says what each system recorded. Context says what the records mean together: a privileged change, an unusual export and a source that has gone quiet are one story, and no system that sees only one of them can tell it.
COMPLEXITY GROWS FASTER THAN MANAGEMENT VISIBILITY.
Growth multiplies the people, services, agents and credentials that can act for the company. Governance has to stay one layer, with one decision path.
The estate keeps growing. The layer that watches it holds steady.
For scaling companiesScaling, and the maturity journey
For scaling companies
COMPLEXITY GROWS FASTER THAN MANAGEMENT VISIBILITY.
Scaling creates more users, more services, more vendors, more sites, more agents, more automation, more credentials, more data, more integrations and more administrative paths.
Leadership remains responsible for all of it.
The scale of the company
Identities and services multiply. Governance stays centralized.
Move the control. The field grows; the governing layer, its five-step decision path and the record it keeps do not. That is the difference between scaling observation and scaling headcount.
Illustrative demonstration
Governing layerThe same at every scale
Identity
Scope
Policy
Authority
Decision
People employees, contractors
Services service accounts, CI/CD
Credentials API keys, temporary credentials
Agents bots, AI agents
Vendors third-party connections
Scale 1 of 3: Team
A few people, a few services.
Leadership can still see most of it directly. One governing layer, one decision path.
Scale 2 of 3: Company
More people, services, vendors and integrations.
Each arrives with credentials of its own. The governing layer is the same one, and the decision path has the same five steps.
Scale 3 of 3: Operation
People, machines, vendors and agents, across sites and environments.
Leadership is still responsible for all of it. Still one governing layer, still one decision path.
One governing layer. One decision path. One record. One picture.
The identity explosion
Every identity carries some authority.
Employees, contractors, service accounts, API keys, CI/CD, bots, AI agents, vendors, temporary credentials and privileged sessions. As identities and machine authority multiply, the boundary that decides matters more, not less.
Illustrative demonstration
How Guard decides
Kinds of identity a growing business must account for
One of six outcomes: allowed, constrained, challenged, revoked, contained or refused. Each is recorded in the Ledger with its justification.
Every request carries an identity: a person, a service, a credential, an agent. Guard begins with who is asking, before anything else.
The identity is resolved to a scope: the tenant and workspace it belongs to. A credential scoped to one workspace cannot read, write or enumerate across another.
Policy is evaluated on the request path, not reconciled afterwards. A decision that cannot be evaluated is refused rather than allowed pending review.
The strength of what Sentinel actually witnessed sets a ceiling on how hard it may respond. Anything stronger than that ceiling needs a named person.
One of six outcomes: allowed, constrained, challenged, revoked, contained or refused. Each is recorded in the Ledger with its justification.
Simulated states, written to show how the system behaves. Nothing here reads or changes a live system.
The path shown is the one Guard applies to a request at Sentinel’s own boundary. What Guard does today: tenant isolation at the data path, graded operator authority, session revocation, attestation-bounded response and provider-confirmed network containment. Guard’s network enforcement is part of the Professional and Enterprise plans; the pricing page states what every plan includes. Sentinel Guard.
What scales
Observation and governance, without a matching increase in people watching.
Sentinel provides a layer capable of scaling observation and governance without requiring the same linear increase in human monitoring. The people a company adds are for judgment, not for watching every new path by hand.
FROM FOUNDER TO ENTERPRISE.ONE LOOP THAT GROWS WITH THE COMPANY.
Every company passes through the same stages, and it should not have to change what protects it at each one. The Sentinel loop is the same from the first day; what grows is how much of the company it reaches.
Choose the stage your company is at to see what changes, what Sentinel does there and where to go next.
Illustrative demonstration
Inner ring: the Sentinel loop. Around it: the people, systems and services of the company at this stage.
The companyThe founder sees almost everything directly: the email, the domain, the payments, the cloud console.
What Sentinel doesAn independent layer begins watching the systems around the founder, and the record begins.
Not more alerts. Context, governance and a record.
AI and autonomous systems
WHO WATCHES THE MACHINES ACTING FOR YOUR COMPANY?
Automation acts faster than people, so its mistakes propagate faster too. Sentinel is built around a ceiling on what any automated action may do on the strength of its evidence, a person above the ceiling, and a record of what happened.
Illustrative: the principle of authority bounds, shown as a picture. It is not a live integration with any agent.
How Sentinel thinksCorrelation, information, the machines, alerts to decisions, existing tools
The correlation story
ANY ONE SIGNAL MAY BE INNOCENT.TOGETHER, THEY REQUIRE ATTENTION.
A business rarely meets one dramatic event. It meets five small ones, in five different systems, each with a perfectly good explanation. Sentinel exists to see the relationship.
Watch the signals arrive one at a time. Then watch what changes when they are read together. Let it play, or step through it.
Illustrative demonstration
Five separate signals, one incident
01 From identity
Unusual privileged session
Could simply be An administrator working from a new place or device.
Not yet seen
02 From protected information
Large protected download
Could simply be A scheduled migration, an audit request or a handover.
Not yet seen
03 From network
Unknown network source
Could simply be A mobile network, a hotel connection or a new VPN.
Not yet seen
04 From export tooling
Rapid export activity
Could simply be A reporting job or month-end work.
Not yet seen
05 From telemetry
Telemetry source goes quiet
Could simply be A maintenance window, a restart or a network fault.
The engine that raises defensive posture as signals accumulate is being built. It is not live.
Step 8 of 8
A decision, a governed action, a record
A person is shown one picture, with the evidence already preserved. Guard acts only within its ceiling, and anything beyond it is theirs to decide. What happens next is recorded, so it can be verified afterwards.
A privileged account begins a session in a way that differs from its own history. It could simply be an administrator working from a new place or device. Alone, it is noted and nothing more.
A large amount of protected material is downloaded. It could be a scheduled migration, an audit request or a handover. Alone, it is routine.
The activity arrives from a network source not seen before for this account. A mobile network, a hotel connection or a new VPN would look the same. Sentinel records the source. It does not say where the person is, or who.
Records are exported in quick succession. A reporting job or month-end work would look the same.
A source that normally reports stops reporting. A maintenance window or a fault would look the same. Sentinel does not read silence as health: loss of visibility is itself a security event.
Five signals, from five systems, around one account in one window of time. Each has an innocent explanation. Together they may describe information leaving through a valid session while the means of seeing it weakens. That relationship is what Sentinel exists to see. It says this requires attention. It does not say it is an attack.
Watchtower sees visibility. Shield sees information movement. Guard governs the response. Evidence preserves the chain. Command shows one picture. Gambit may increase defensive scrutiny.
A person is shown one picture, with the evidence already preserved. Guard acts only within its ceiling, and anything beyond it is theirs to decide. What happens next is recorded, so it can be verified afterwards.
A conceptual example with no customer, company or event in it. Simulated states, written to show how the systems divide the work. Nothing here reads or changes a live system.
What is built. Each system carries the status its own page gives it. Joining signals from several systems into one picture is the work of Command and of Gambit, and Gambit, the part that raises defensive scrutiny as signals accumulate, is being built. It is not live.
SENTINEL TURNS SIGNALS INTO DECISIONS, DECISIONS INTO GOVERNED ACTION, AND ACTION INTO A VERIFIABLE RECORD.
SignalsWatchtower and Shield observe
DecisionsCommand shows one picture; a person decides what is theirs
Governed actionGuard, within its ceiling
Verifiable recordEvidence: Ledger, Trace, Pulse
The information problem
NOT EVERY THREAT ENTERSTHROUGH A BROKEN PASSWORD.
Information can leave through valid accounts: employees, contractors, vendors, a legitimate session that has been compromised, or a careless action. A valid login does not automatically mean every action after it is safe.
This is where Shield matters. It does not accuse a legitimate user. It records what happened to protected information, with the basis for every fact, so that a question can be answered with evidence.
Information can leave through
Employees
Contractors
Vendors
A compromised legitimate session
A careless action
None of them needs a broken password. Each already has a way in. The question is what happens to the information afterwards.
Illustrative demonstration
A protected document, from release to record
UnclassifiedConfidential
Servedonce
Bound to a session
LedgerTracePulse
Classified
Classification recorded: confidential.
Watermarked
Watermark record: bound to this render, digest only.
Served once
Grant: authorised first, consumed once.
Attributed to a session
Session recorded. No identity inferred from an address.
Preserved as evidence
Evidence preserved. Audit state stated, never assumed.
Step 5 of 5
Preserved as evidence
The download and its watermark are preserved as Evidence, with the audit state stated truthfully: recorded, pending, failed, or no evidence. Later, a copy found elsewhere can be matched to the render it came from by re-deriving its mark. Evidence storage is verified per deployment.
The document carries a classification: public, internal, confidential or restricted. It is consulted at the API boundary, and it decides who may read the document, whether a mark is rendered and whether a download can be granted at all.
A mark is derived from the tenant, site, session, content and policy and embedded in the copy that is rendered. It is a keyed digest, never a person's identity. Where a watermark policy applies, the copy is marked or the download is refused. Today marking is built for text content; other formats are not yet marked.
The download is authorised to a real, signed-in session before it exists, and the authorisation is written to the Ledger first. At the moment of release every gate is checked again, and the grant is consumed exactly once. A second attempt is refused, and the refusal is recorded.
The release is bound to the session that received it, not to a guess about a person. A watermark attributes; it does not prevent. A network address is a source, never a person and never a place.
The download and its watermark are preserved as Evidence, with the audit state stated truthfully: recorded, pending, failed, or no evidence. Later, a copy found elsewhere can be matched to the render it came from by re-deriving its mark. Evidence storage is verified per deployment.
One document followed through the stages Shield records. Simulated, written to show the order of the stages. Nothing here reads, serves or marks a real document.
What Shield records, and what it never says
Shield records
Which classification a document carries.
Which session a download was authorised to, and that it was served once.
The mark bound to each render, as a digest and never as an identity.
Copy and print initiated inside protected content, where the protection tier permits collecting it.
Extraction conclusions, each with its basis and never stronger than its weakest basis.
Shield never says
That a screenshot was taken. A browser cannot see an operating-system screen capture.
Where a person is, or who they are, from a network address. An address is a source.
That a watermark prevents copying. A watermark attributes.
That a withheld copy or print option is a boundary. The content is already on the machine.
That anyone intended anything. Extraction is a fact about bytes, never about intent.
What exists today
Shield is built in part. This is the plain account of which parts, read from the platform’s own record of what it ships.
Built
Classification, and the policy resolver that decides from it
Protected downloads, authorised before they exist and served once
Watermark records bound to a render
Extraction and scraping conclusions, with their basis
Preservation of Shield records as Evidence
Built in part
Watermarking: text content only
Copy and print signals: only where the protection tier permits, and a page's own report is never proof
Session attribution: only where tier and classification authorise it
Being built
Watermarks beyond text
Sentinel Gambit, the adaptive response to what Shield records
Endpoint-attested capture, which needs an enrolled managed endpoint
AI and autonomous systems
WHO WATCHES THE MACHINES ACTING FOR YOUR COMPANY?
Businesses increasingly give software and AI agents authority: to read, write, deploy, communicate, analyze, purchase, run workflows, move information and make decisions. Automation acts faster than people. Its mistakes propagate faster too.
A company that gives authority to a machine needs what it needs from any delegate: a ceiling on what it may do, a standard for the evidence behind an action, a person to ask when the ceiling is reached, and a record of what happened. Sentinel is built around that discipline.
Illustrative demonstration
An authority ceiling, applied
Try
Capped to challenge
A site reported observation may justify at most challenge. Block requires stronger evidence or an authorised person.
A named personAnything above the ceiling, and anything only a person may choose.Can authorise more
BlockRefuse the source entirely.Requested
Revoke sessionEnd a session that was already open.Above the ceiling
RestrictNarrow what the source may do.Above the ceiling
ThrottleLimit how often the source may act.Above the ceiling
ChallengeAsk the source to prove itself before it continues.AppliedCeiling for this evidence
ObserveWatch and record. Nothing changes for anyone.
Wanted
Block
Applied
Challenge
Decided by
Guard, autonomously. Nobody was asked.
Written to
The Ledger, with the reason and the evidence basis. What was wanted is kept beside what was done.
An imagined request, run through the rule Guard enforces. Nothing here reads or changes a live system.
What is built. Guard’s ceiling is built and tested: it is the rule for what an automated action may do without a person. Today an operator makes every network rule, and Sentinel does not yet act against a source on its own. Extending the same ceiling, escalation and record to the agents and automations a company deploys is the direction. It is not a shipped capability, and this page does not present it as one.
The rules, in full
The strength of the evidence sets a ceiling. Repetition never raises it: ten thousand opinions from a browser are still a browser’s opinion.
Most severe action without a person
Evidence
Ceiling and what the evidence means
Inferred
ChallengeSentinel inferred it. Nothing reported it.
Site reported
ChallengeA page's own agent said so. A browser is something an attacker controls.
Server observed
ThrottleSentinel's own server saw it, against its own clock.
Edge observed
BlockProved cryptographically at the edge. No edge attestation path is built, so nothing can carry this level today.
A range of addresses contains addresses nobody enumerated. A person must see the range first. A rule with no expiry outlives everyone who remembers why it exists. A person must choose that. A rule for every site reaches the whole estate. The blast radius is the estate. Where any part of a conclusion is an inference, the ceiling is a challenge, whatever else sits beside it.
What a company is entitled to ask of an autonomous system
Authority ceilings
The strength of the evidence sets the most severe action Sentinel may take on its own. Evidence that is only inferred, or only reported by a page, stops at a challenge.
Built
Evidence strength
Every observation says how it is known: inferred, site reported, server observed or edge observed. Repetition never converts a weak signal into a strong one.
Built
Policy
A tenant's policy may tighten what is allowed. Nothing loosens a ceiling.
Built in part
Auditability
Every decision is written to the Ledger with its reason, its policy, the strength of its evidence and who or what decided. What was wanted is kept beside what was done.
Built
Human escalation
A range of addresses, a rule with no expiry and a rule for every site always need a named person, whatever the evidence. The owner's allowlist is checked before authority, and the role that authorised an action is recorded at the time.
Built
Session and action constraints
A session can be revoked. Safe Mode, engaged by an operator, holds Sentinel's own high-risk operations within a scope while evidence capture and incident response keep running.
Built
Built today, and the direction
Built today
Guard's evidence ceiling is built and tested. It is the one rule for what an automated action may do without a person.
Every network rule today is made by an operator. The actions only an owner may take, a range, no expiry or every site, are the ones the ceiling reserves for a person.
Every decision is recorded with its reason, the strength of its evidence and who decided.
Sentinel does not yet act against a source on its own, and no autonomous block is possible: the level of evidence that would allow one does not exist yet.
The direction, not yet shipped
Bringing the actions of the agents and automations a company deploys under the same ceiling, escalation and record.
Treating an agent's authority as something granted, bounded and revocable, with its own history in Evidence.
No integration with a company's own AI agents is shipped. The demonstration above governs Sentinel's own responses.
Alert fatigue
FROM ALERTS TO DECISIONS.
A monitoring product that only sends alerts transfers the analytical burden to whoever happens to be awake. Sentinel is built to carry that burden itself: to observe, verify, correlate, prioritize, preserve, respond within authority, and escalate only when necessary.
Illustrative demonstration
From a stream of signals to one decision
A product that only sends alerts
Whoever is awake
Every signal becomes a notification. The analysis is left to whoever happens to be awake.
Sentinel
01Observe
02Verify
03Correlate
04Prioritize
FirstNextLater
05Preserve
Record
06Respond within authority
ActionWithin authority
07Escalate only when necessary
A personOnly if needed
One decision. One record.
A governed action, with the record of why it was taken.
Step 7 of 7
Escalate only when necessary
A person is brought in when that authority is reached. What reaches them is one decision with its record, not a stream to sort.
Signals arrive from everything Sentinel watches. At this point they are only signals: some meaningful, many not.
Each signal is checked: is it real, and does it come from a source Sentinel trusts? What cannot be verified is held at a lower standing instead of being treated as fact.
Signals about the same account, session or change are related to one another. Many signals become a few situations.
The situations are ordered by consequence, not by the order they arrived in. One needs attention first.
Before anything is changed, the chain of what was seen is written down, so the response cannot erase the evidence for it.
Sentinel acts within the authority it has been given, and no further. Guard’s ceilings decide what it may do on its own.
A person is brought in when that authority is reached. What reaches them is one decision with its record, not a stream to sort.
Simulated states, written to show how the system behaves. Nothing here reads or changes a live system.
SENTINEL TURNS SIGNALS INTO DECISIONS, DECISIONS INTO GOVERNED ACTION, AND ACTION INTO A VERIFIABLE RECORD.
Existing tools
WHY NOT JUST USE EXISTING TOOLS?
You may already run excellent tools: an identity provider, email protection, a cloud platform, endpoint agents, a payment system, a ticketing queue. None of that is the problem. Sentinel answers a different question: what do all of it, together, mean?
Illustrative demonstration
Each tool knows part of the organization
Identity
Email
Cloud
Endpoints
Network
Payments
Sentinel
Identity
Email
Cloud
Endpoints
Network
Payments
Sentinel
Each dot is a person, an account, a system or a vendor.
Correlation
Visibility
One operating picture
Governance
Evidence
Continuity
Step 5 of 5
One operating picture
On that picture Sentinel governs the response within authority, keeps what happened as evidence and keeps the organization recoverable. One place to see, decide and verify.
An identity provider sees who signed in and with what privilege, and it does that well. It does not see what the account then did in the payment system, the cloud console or the repositories.
Email protection sees mail. A cloud platform sees workloads. Endpoint tools see devices. Each is excellent at its job, and each stops at its own edge.
Together the windows cover much of the business. But each reports in its own language, to its own console, on its own schedule, and nothing connects what one sees to what another sees.
Sentinel is the layer that relates what is known about one account, one session or one change across the windows it observes. That relationship is the picture no single tool has.
On that picture Sentinel governs the response within authority, keeps what happened as evidence and keeps the organization recoverable. One place to see, decide and verify.
Illustrative. The windows are categories of tool, not any product's coverage, and the shapes are not to scale. Nothing here reads a live system.
SENTINEL DOES NOT NEED EVERY OTHER TOOL TO BE BAD.IT EXISTS BECAUSE NO SINGLE TOOL SEES THE WHOLE BUSINESS.
What Sentinel does not doSentinel does not replace these tools and does not see through them. It relates what it is connected to, and it shows plainly what it cannot see.
04
Part 04 of 06
The system
Specialized systems. One continuous loop.
Command unites the picture
One picture of every source, and how each one is known.
Specialized systems, one continuous loop. This is the surface a person sees it on.
Illustrative demonstration
Sentinel CommandOperational picturedegraded
AXIS OMNIA stops reporting.
Expected telemetry from AXIS OMNIA is no longer arriving. Sentinel does not read silence as health: the gap is raised as a security condition and the investigation widens to ask why this source went quiet now.
Sources
AXIS 7DserviceReporting
AXIS COREserviceVerified
AXIS OMNIAserviceSilent
CLOUDFLARE EDGEedgeVerified
SENTINELcontrol planeVerified
D1 INDEXevent indexReporting
EVIDENCE STOREraw evidenceReporting
COLOSSEUMcontinuityReporting
Events
TELEMETRY_SOURCE_SILENTcriticalWATCHTOWER · checked at gateway · evidence recording
COVERAGE_BELOW_EXPECTEDhighWATCHTOWER · checked at gateway · evidence recording
What Sentinel does with silence
A source that stops reporting is not read as healthy. The gap is a finding, and it is worked like any other.
Command unites the picture: every source, how it is known, and what changed. Illustrative demonstration with fabricated identifiers; nothing here reads a live system.
The systemThe Sentinel system, the chessboard
The Sentinel system
FIVE SPECIALIZED SYSTEMS.ONE CONTINUOUS PROTECTION LOOP.
Sentinel is not one tool watching one thing. It is an operating loop: observe, verify, detect, protect, preserve, recover. Each system owns part of that loop, and each answers a question the business would otherwise answer by hand.
Illustrative demonstration
The Sentinel operating loop
01ObserveWatchtowerFirst Light
02VerifyFirst Light
03DetectWatchtowerShield
04ProtectGuardShield
05PreserveEvidence
06RecoverContinuity
CommandRange validates the loopFirst LightShieldGambitPresence
Each role answers a question the business would otherwise answer by hand: what is happening, who may act, whether the control holds, what happened and how to recover. Command holds the picture over all of it.
A system that stops reporting is not a system that is fine, so Watchtower compares what should be reporting with what is, and treats the gap as a finding.
Growth multiplies what can act on the company, so Guard decides authority at the boundary instead of trusting it to hold by habit.
A control that has never been tested is an assumption, so Range exercises Guard, recovery and Sentinel's own control plane against defensive scenarios before an incident does.
The most expensive question after an incident is often what actually happened, so Evidence keeps the record: Ledger for what happened, Trace for how, Pulse for what is true now.
Prevention is never absolute, so the organization needs a way back: Continuity requests preservation, records the outcome and holds an incident open until recovery is verified.
Information can leave through a valid account, so Shield classifies what matters, marks it when it is released and watches how it leaves.
A static defense waits for the next rule violation, so Gambit raises scrutiny in proportion to what Sentinel can establish, and it never retaliates.
Leadership should not assemble the picture from separate tools, so Command shows one operating view where a person sees, understands, authorizes, acts and verifies.
Customers should be able to see that protection is active, so Presence lets a page show that Sentinel is reporting on it, and only a page registered with Sentinel can.
Every conclusion depends on its signals, so First Light admits each observation through one trust boundary and records how it is known.
Each role answers a question the business would otherwise answer by hand: what is happening, who may act, whether the control holds, what happened and how to recover. Command holds the picture over all of it.
Illustrative demonstration of where each system sits on the operating loop. Nothing here reads or changes a live system. What is built today is stated on each system's own page.
Ten roles, one line each
Each line links to the system’s own page, where its boundaries and its behaviour when it fails are stated. Where a system is only partly built, its tag says so.
WATCHTOWER SEES.
A system that stops reporting is not a system that is fine, so Watchtower compares what should be reporting with what is, and treats the gap as a finding.
A control that has never been tested is an assumption, so Range exercises Guard, recovery and Sentinel's own control plane against defensive scenarios before an incident does.
The most expensive question after an incident is often what actually happened, so Evidence keeps the record: Ledger for what happened, Trace for how, Pulse for what is true now.
Prevention is never absolute, so the organization needs a way back: Continuity requests preservation, records the outcome and holds an incident open until recovery is verified.
Leadership should not assemble the picture from separate tools, so Command shows one operating view where a person sees, understands, authorizes, acts and verifies.
Customers should be able to see that protection is active, so Presence lets a page show that Sentinel is reporting on it, and only a page registered with Sentinel can.
A static defense waits for the next rule violation. Sentinel can progressively change its defensive posture as evidence accumulates: Observe. Attribute. Watermark. Challenge. Constrain. Preserve. Verify. Escalate.
Posture steps up only as the evidence beside it does, never on repetition alone, and it steps back down when the evidence lapses.
What is built. Sentinel Gambit, the engine that moves posture as evidence accumulates, is being built. It is not live. The measures it moves between exist in part today, and every rank below says which.
Illustrative demonstration
Their move, Sentinel's answer
01Observe
02Attribute
03Watermark
04Challenge
05Constrain
06Preserve
07Verify
08Escalate
Sentinel’s postureEscalate
They moveSentinel moves with them
01Not yet played
They
A valid session reads protected documents within its usual pattern.
Server observed Sentinel's own server saw it.
Sentinel
Observe
Sentinel records the activity with its basis. Nothing changes for the person using the session.
BuiltShield records signals with their basis, where the protection tier permits.
02Not yet played
They
The session opens protected documents in a broad sweep.
Server observed Sentinel's own server saw it.
Sentinel
Attribute
Every record is bound to this session. A session is not a person, and a network address is not a place.
Built in partSession attribution exists where the tier and the classification authorise it.
03Not yet played
They
The session asks for a protected download.
Server observed Sentinel's own server saw it.
Sentinel
Watermark
The copy is served once and carries a mark bound to this render. A watermark attributes. It does not prevent.
Built in partMarking is built for text content; other formats are being built.
04Not yet played
They
The browser reports activity consistent with automation or an attempted capture.
Site reported A page's own report. A browser is something an attacker controls.
Sentinel
Challenge
The session is asked to prove itself again. This is as far as a browser's report can take Sentinel, however often it repeats.
Held at the ceiling: a page's own report may justify at most a challenge.
Built in partGuard can request a challenge for a network source through a provider, and a rule is active only when the provider confirms it. Challenging a session itself is part of Gambit.
05Not yet played
They
Protected content keeps leaving, and the server observes it.
Server observed Sentinel's own server saw it, so the evidence is its own.
Sentinel
Constrain
What the session may do is narrowed within Guard's ceiling: shorter-lived grants, restricted downloads.
Being builtAdaptive constraint is the part of Sentinel Gambit that is being built.
06Not yet played
They
The activity continues from a different network source.
Server observed Sentinel's own server saw it.
Sentinel
Preserve
The chain is preserved as evidence before anyone remediates: the signals, the decisions and the grants, each with its basis.
BuiltShield records are preserved as Evidence. Evidence storage is verified per deployment.
07Not yet played
They
The session goes quiet.
Server observed Sentinel's own server saw it.
Sentinel
Verify
Sentinel checks that its response took effect and that its own telemetry is still arriving. A rule is active only when the provider confirms it.
Built in partProvider state is read back, never assumed. Watchtower reports when visibility is lost.
08Not yet played
They
The pattern remains unexplained.
Server observed Sentinel's own server saw it.
Sentinel
Escalate
A person is asked, with the evidence in front of them. Containment beyond the ceiling is a human decision.
BuiltGuard escalates to a named person, and authority is graded by role.
Step 8 of 8
08. Escalate
The pattern remains unexplained. Sentinel escalates: a person is asked, with the evidence in front of them. Containment beyond the ceiling is a human decision. Stepping back down is a decision too, by an operator or a recorded timeout, never the engine's alone.
A valid session reads protected documents within its usual pattern. Sentinel observes: it records the activity with its basis, and nothing changes for the person using the session.
The session opens protected documents in a broad sweep. Sentinel attributes: every record is bound to this session. A session is not a person, and a network address is not a place.
The session asks for a protected download. Sentinel watermarks: the copy is served once and carries a mark bound to this render. A watermark attributes. It does not prevent.
The browser reports activity consistent with automation or an attempted capture. Sentinel challenges: the session is asked to prove itself again. This is as far as a browser's report can take it, however often it repeats. The ceiling holds here.
Protected content keeps leaving, and the server observes it. Now the evidence is Sentinel's own, so Sentinel may constrain: within Guard's ceiling it narrows what the session may do, with shorter-lived grants and restricted downloads. This is the part of Gambit that is being built.
The activity continues from a different network source. Sentinel preserves: the chain is kept as evidence before anyone remediates, with the signals, the decisions and the grants, each with its basis.
The session goes quiet. Sentinel verifies: it checks that its response took effect and that its own telemetry is still arriving. A rule is active only when the provider confirms it.
The pattern remains unexplained. Sentinel escalates: a person is asked, with the evidence in front of them. Containment beyond the ceiling is a human decision. Stepping back down is a decision too, by an operator or a recorded timeout, never the engine's alone.
An imagined scenario with no customer and no person in it: a valid session begins to move protected information. Simulated, written to show how posture follows evidence. Nothing here reads or changes a live system.
SENTINEL DOES NOT HACK BACK.
Nothing here touches a system outside the protected estate, and nothing attacks whoever opens a decoy. Sentinel wins through what it can see and prove, and what it can do inside its own estate.
Visibility
Policy
Deception within the protected estateBeing built Opt-in, and being built.
Attribution
Containment
Evidence
Verification
A sentence is not said unless its fact is recorded
Sentinel can say more when it can prove more. A browser cannot see a screen capture, so a browser signal never produces the words below. Switch the facts on and off, and watch the words follow them.
Illustrative demonstration
When Sentinel may say more
Claim Sentinel can stand behindSCREEN CAPTURE DETECTED
I SEE YOU.
This activity has been recorded.
Your session has been flagged for review.
A demonstration of a sufficiently attested capture. The attestation is a placeholder: no real provider, device or person is involved, and nothing here reads or changes a live system.
What a complete attestation contains
Provider kind
A managed endpoint agent, enterprise browser extension, managed desktop client, MDM, EDR or platform API. A web page is none of these.
Provider
The named product that reported it.
Device
Which enrolled device it came from.
Session
Which session it belongs to.
Observed at
When the provider says it happened.
Policy
The tenant policy the endpoint reports under.
Basis
How the provider knows.
What the words never say
Where you are. A network address is a source, never a place.
Who you are. Sentinel does not name a person.
That a screenshot was taken, unless a trusted endpoint attested a capture. A browser cannot see one.
That anyone intended anything, or did anything wrong.
Anything threatening, and nothing about retaliation. Sentinel does not hack back.
05
Part 05 of 06
The record and the way back
What happened, who authorized it, and how the company recovered.
Business continuity
PROTECTION IS NOT ONLY ABOUT PREVENTING THE INCIDENT.IT IS ABOUT ENSURING THE ORGANIZATION SURVIVES IT.
Perfect prevention is not a credible promise. Sentinel detects, contains and verifies. AXIS COLOSSEUM, a separate system, preserves and recovers.
Illustrative: how an organization survives an incident. Sentinel verifies; COLOSSEUM preserves and recovers.
The record and the way backThe operational record, continuity, the weekly report, visible trust
Evidence and accountability
EVERY MAJOR FUNCTION HAS A SYSTEM OF RECORD.WHERE IS THE ONE FOR PROTECTION?
A business keeps a customer record, a resource record, a financial record and a people record. Ask where the authoritative record of its incidents, its security decisions, its containment, its visibility failures, its evidence, its protective actions, its recoveries and the trust in its own systems lives, and the answer is often fragments: logs in one place, tickets in another, memory in a few heads.
Illustrative demonstration
CRM
Customer record
Who is the customer, and what do we know?
ERP
Resource record
What do we hold, and where is it?
Accounting
Financial record
What came in, and what went out?
HRIS
People record
Who works here, and in what role?
Sentinel
Protection record
How does the organization protect itself, and what happened when it had to?
Incidents
Security decisions
Containment
Visibility failures
Evidence
Protective actions
Recovery
System trust
Often scattered across
Logs
Tickets
Chat threads
Inboxes
Memory
THE OPERATIONAL RECORD OF HOW THE ORGANIZATION PROTECTS ITSELF.
That is the direction Sentinel is built toward: not another place to look, but the place where the answer is already written down.
One incident. Three views of the same record.
Illustrative demonstration
One incident
Ledger
What happened
Durable history, appended in order: the detection, each decision, each action, each hand-off.
Answers the question
Trace
How it happened
The causal sequence, reconstructed across services, policy evaluations and boundaries.
Answers the question
Pulse
What is true now
Current operational state, derived from the record: coverage, posture, active enforcement and open incidents.
Answers the question
A leader asks
Answered from Ledger
The Ledger holds the ordered history of the incident: what was detected, what was decided and what was done, in the order it happened.
What actually happened?
The Ledger holds the ordered history of the incident: what was detected, what was decided and what was done, in the order it happened.
How did it happen?
Trace follows the incident across services, policy evaluations and boundaries, so an investigator reads a sequence instead of inferring one.
What is true right now?
Pulse states the current operational state, derived from the record: what is covered, what is contained, what is being enforced and what is still open.
Who authorized the response?
Guard decisions are written with the change, and an authorized response is recorded with the authority it was taken under.
What was kept, and is it intact?
Raw evidence is kept apart from its searchable index, with its digest and provenance, and each object is reported as verified, unverified, mismatched or missing. Where storage is not verified, Sentinel refuses evidence rather than pretend.
Did the recovery work?
An incident cannot be closed while recovery verification is open, and the outcome is written to the Ledger. Pulse shows the state the organization is in now.
What the record is notIt is the record Sentinel holds of what it observed and decided, within today’s architecture. The Ledger is append-oriented and hash-chained, so a silent edit to one entry becomes detectable; it has no external anchor. It makes no claim about legal standing and does not certify compliance with any regime; that assessment belongs to counsel and auditors. Raw evidence objects need storage that is verified per deployment. That storage is not included in any plan, and a deployment without it refuses evidence rather than pretend.
Degraded by design. Not broken.Safe Mode boundary, lifted once the state was verified.
SENTINEL PROTECTS AND VERIFIES.
COLOSSEUM PRESERVES AND RECOVERS.
01Detection
02Preservation request↓ COLOSSEUMCOLOSSEUM →
03Verified restore point↑ SENTINEL← SENTINEL
04Containment
05Recovery
06Integrity verification
07Return to operation
Step 7 of 7
Return to operation
With the state verified, an operator lifts Safe Mode and the organization expands back out. Every step is in the record.
Sentinel notices a condition that needs a response, and an operator declares the incident. Prevention is never absolute; this is the moment the rest of the process exists for.
At the moment of detection, before any remediation could overwrite state, Sentinel asks for critical state to be preserved. The request is tied to the incident.
AXIS COLOSSEUM, a separate system, holds the restore point and confirms it. Sentinel records the confirmation to the Ledger.
Guard narrows what the affected identities and sessions can reach. An operator with authority can engage Safe Mode: the organization contracts to a protected operating core. Degraded by design, not broken.
COLOSSEUM executes recovery from the restore point. Sentinel never performs backup, snapshot or restore.
Sentinel verifies the recovered state. An incident cannot be closed while a recovery-verification task is still open.
With the state verified, an operator lifts Safe Mode and the organization expands back out. Every step is in the record.
Simulated states, written to show how the system behaves. Nothing here reads or changes a live system.
PROTECTION IS NOT ONLY ABOUT PREVENTING THE INCIDENT.IT IS ABOUT ENSURING THE ORGANIZATION SURVIVES IT.
SENTINEL PROTECTS AND VERIFIES. COLOSSEUM PRESERVES AND RECOVERS. Two systems, two mandates, never merged.
What exists todayBuilt: preservation requests tied to the incident, outcomes recorded to the Ledger, and a close that is refused while recovery verification is open. Not built: a live COLOSSEUM integration. Today a preservation outcome is an operator’s attestation of what COLOSSEUM reported, and is labelled that way. Safe Mode is engaged by an operator with authority, not by a signal.
Leadership answers for outcomes, not for raw telemetry. Sentinel is designed to turn operational facts into a short weekly report in five plain sections, send the summary to a verified security mailbox, and keep the detail where it can be checked.
Illustrative demonstration
AXIS Sentinel
Weekly leadership report
Example organization · the previous week
Operating stateObserving
What changed
A new administrator was added to the production workspace, and the change was reviewed.
One protected site reported on a different schedule for part of the week, then returned to its usual pattern.
A vendor connection was re-authorized after its credential expired.
What requires attention
A privileged account has not completed its scheduled access review.
One telemetry source stayed quiet for longer than expected and is still being watched.
What Sentinel did
Narrowed a service account’s scope to match its policy.
Preserved the record of the access change and linked it to its review.
Declared no incident and requested no containment.
What remains open
The access review, owned by the operations lead.
The quiet telemetry source, under observation by Watchtower.
Where to verify it
Every line above links to the record behind it in Sentinel Command.
The email is the summary. Command is where it is checked.
Illustrative content for an example organization. Not a real report.
Where the summary goes, and why
Summary
A verified secondary security mailbox
security@, soc@ and sentinel-alerts@ are examples.
Recommended. If the primary mailbox or identity account is ever compromised, security notifications should not depend on that same account alone.
Detail and verification
Sentinel Command
The record behind each line, for an authenticated operator to check.
Not relied on alone
The primary sign-in mailbox
It can receive the report as well, but it is the account most exposed in the very case a secondary mailbox exists for.
A separate mailbox reduces a single point of failure. Sentinel does not promise delivery, including during a compromise.
Written for the people who answer for the business
Founder
CEO
COO
CTO
CISO
Board-level leadership
StatusBeing built. The weekly leadership report and the verified security mailbox are designed and in development. They are not live yet; this page shows the design, not a shipped feature.
Customers cannot read a security architecture. They can see whether a company shows them anything at all. The Sentinel banner is designed to follow the state of a registered deployment: a mark that is true when it is shown, and changes when it stops being true.
Illustrative demonstration
customer-site.example
Sign in
Continue
PROTECTED BY AXIS SENTINEL
Always on guard. Always watching.
MONITORING ACTIVE
An example customer site. The banner is the strip at the foot of the page.
The four honest states
State: Reporting
The banner makes its protection claim, because the deployment is registered and reporting.
What makes it worth trusting
Live
It follows a registered deployment, not a badge that was true on the day it was installed.
Truthful
If Sentinel cannot verify reporting, the banner degrades. If registration is revoked, the claim disappears.
Visible
A customer, a partner or a reviewer can see that the property is being watched, and that someone is accountable for it.
Not a certification
The banner says monitoring is active. It does not say a site cannot be breached, and it is not a statement of compliance.
Where it matters
A young company
Shows the maturity of its protection before it has a long record to point to.
An established company
Reinforces the trust customers already place in it, with something they can see.
A portal, a login or a checkout
Gives visible evidence of active monitoring where a visitor is about to hand something over.
StatusBeing built. The Protection Banner is designed and in development; it is not live yet. The Sentinel Presence mark and the visitor seal it builds on already exist. The states shown here are the design’s own, simulated.
What not knowing costs, who it matters to, and where to begin.
The decisionThe cost of not knowing, why leaders care
The cost of not knowing
THE MOST EXPENSIVE SECURITY QUESTION IS OFTEN:“WHAT ACTUALLY HAPPENED?”
The cost of a serious incident is not one number. It is a set of consequences that arrive together, most of them before anyone can say what happened.
What stops
Operational interruption
Lost access
Lost time
Lost opportunity
What is lost
Lost customer trust
Lost information
Lost intellectual property
Fraud
Reputation damage
What it costs to answer
Recovery cost
Investigation cost
Executive distraction
Regulatory and legal exposure
There is no figure on this page, on purpose. What an incident costs depends on what was reached, how long it went unseen and what can be shown afterward. A number that fits every business would be invented.
The same question. Two starting points.
Illustrative demonstration
Starting from nothing
Question askedRebuilding the recordInvestigatingAnswer
Starting from a record
Question askedInvestigatingAnswer
Starting from nothing
Find where the evidence might be.
Work out what is missing.
Reconcile clocks and accounts from separate systems.
Establish a sequence.
Only then, ask what happened.
Starting from a record
The Ledger already holds what happened, in order.
Trace already follows how it happened.
Pulse already states what is true now.
The investigation begins at the question.
Lengths show the order of effort, not a measured time.
Sentinel exists so the organization is not starting that investigation from nothing.
Why leaders care
THE SAME LAYER.A DIFFERENT REASON FOR EACH LEADER.
Founders, technologists, security leads, operators, finance and the board each answer for a different part of the business. Choose a role to see why Sentinel matters to it.
Showing Founder / CEO. Know what requires attention without reading raw logs.
Founder / CEO
Know what requires attention without reading raw logs.
The question they bringWhat needs my attention this week, and what has already been handled?
No return-on-investment figure is offered here. A figure that fits every business would be invented; what Sentinel reduces is uncertainty, and each role feels that in a different place.