StackSchematic is an application-centric, continuously updated map of everything your organization runs, organized the way the business thinks about it rather than the way infrastructure tooling happens to emit it. Every fact it shows you carries where it came from and how confident you can be in it.
Not because nobody has tried. Because the answers live in a dozen tools that each know one slice, and the document that reconciled them was accurate the week it was written.
Not the list somebody maintained until they left, but the applications that exist, found by looking, and kept current by looking again.
Databases, storage, queues, identity providers, third-party services. Including the dependencies nobody wrote down.
A diagram says an application calls a service. StackSchematic can tell you whether anything has ever been observed doing so.
What must be restored? What can be rebuilt? What belongs to somebody else? And has the recovery ever been tested?
This is the idea from which the rest of the product follows. Most tools present a guess and a measurement identically. StackSchematic keeps them apart, permanently, in the data model rather than in a footnote.
An inference is never silently promoted into a declared fact. Turning one into a statement your organization stands behind takes an explicit, attributable human action, and that action stays on the record.
The conflicts are the valuable part. “Declared but never observed” is a dependency somebody believes in that nothing uses. “Observed but never declared” is a dependency nobody knew about until it broke. Most tools resolve these away. StackSchematic reports them.
Each kind of evidence has exactly one product that owns it. That is not an implementation detail; it is what makes “never present a guess as a fact” something the product's structure guarantees, not a rule everyone has to remember to follow. The eighth, Guide, owns none — an optional assistant that reads all of them and asserts nothing of its own.
The application catalog. What exists? Who owns it? What is it for? How critical is it? The layer a person writes, and the one an organization is judged on.
The technology graph: entities, relationships, provenance, confidence. The one place every kind of evidence meets, and where conflicts between them surface instead of being averaged away.
Automated discovery from a fleet of agents, each reporting what it can see from where it runs: cloud accounts, clusters, hosts, identities, repositories, on-premises assets.
Runtime truth. What is actually being used? What answers a check? How recently? The difference between an application that works and one nobody has looked at.
The visual, application-centric picture: an application and everything beneath it, with each relationship showing what it rests on. Reads and never writes, so a presentation choice can never become an assertion.
Attaching, configuring and monitoring the sources evidence comes from, so adding one never means touching how the estate is reconciled.
Whether you could get an application back. What must be restored? What can be recreated? What belongs to another team and is merely verified? All assembled into an ordered recovery plan.
The optional assistant over the whole blueprint. It reads every product through one boundary and owns no evidence of its own. It is the only member of the family that asserts nothing, and the only one you can switch off.
Guide answers questions across the whole blueprint — “what would we need to restore if this were lost?”, “which systems have undocumented recovery procedures?”, “what is not protected?” — and every answer it gives carries the same evidence labels the rest of the product uses.
Anything that changes your estate or reaches real infrastructure requires an explicit human approval, recorded against the person who gave it. No assistant, now or later, launches a scan or alters the graph on its own initiative.
StackSchematic is fully functional with Guide switched off, and nothing else depends on it. Organizations that do not want an assistant in the loop lose nothing else.
Asked something the estate cannot answer, Guide says so, including when a search ran without the full capability behind it. An assistant that always produces an answer is an assistant you cannot use for anything that matters.
Most organizations start with nothing: no catalog, no CMDB, often no reliable service inventory. So the usual order runs backwards, and that is the intended path rather than a fallback.
Agents report what they can reach from where they run. Nothing is assumed about parts of the network nobody can see, and that gap is stated rather than left blank.
The graph proposes what belongs to what, and labels every proposal as its own inference, including the ones it is confident about.
A person accepts, corrects or rejects. That act is what turns a proposal into a declared fact, and it stays attributable afterwards.
Discovery runs again. Staleness is itself a recorded fact, so a blueprint that has stopped being maintained says so rather than ageing silently.
Where a fact came from travels with it through every layer, to the screen and to the assistant. A surface that cannot express it is treated as a defect in the surface.
“We looked and found nothing” and “we could not look” are different answers, and StackSchematic never collapses them into one.
Anything that changes what the system asserts, or reaches your infrastructure, is traceable to a person and a reason.