/ technology comprehension infrastructure
Understand your technology.Then make it work.
Inferrex builds a shared understanding of systems, APIs, data and the relationships between them. Integration, automation, intelligence, governance, migration and AI are applications of that understanding, not separate products.
- 234,373
- Vendors catalogued
- 7,194,438
- Fields comprehended
- 1,733
- Canonical keys
Self-healing
Stripe API
Customer · monitored for change
- Stable
- Change detected
- Re-inferred
- Remapped
time · created at
Mapping healthy, data flowing.
/ the problem
Technology can be connected without being understood.
Most estates are understood through documentation, point-to-point integrations, tribal knowledge, spreadsheets, manual mapping, repeated discovery and fragmented tooling. Each of those is a way of describing a system to a human being. None of them is a way of making a system legible to software.
So organisations connect systems without ever building a durable understanding of what those systems mean. The connection works until the provider changes. The understanding was never written down anywhere a machine could read it, so it is rebuilt, by hand, every time.
/ the model
Systems, comprehended once, and every outcome read off the result.
A system arrives as a specification, a schema or an endpoint. Inferrex resolves its structure, infers what each element means, expresses that meaning against a canonical model, then relates it to every other system already comprehended.
RAW SYSTEM → STRUCTURE → MEANING → CANONICAL MODEL → RELATIONSHIPS → OUTCOMES
Deterministic inference does the work wherever the answer is determinable; AI is reserved for the residual uncertainty where it is actually needed. The result is not a description of a system. It is a model of one, and the outcomes below are read off it.
/ outcomes
One understanding. Six applications.
Not six products. Six things that become possible once the understanding exists.
Integrate
Connect systems without rebuilding the same understanding every time. A contact in one system, a person in another and a customer record in a third resolve to the same canonical concept, so the mapping is derived rather than authored.
How integration worksAutomate
Turn understood relationships into workflows that survive change. When a provider alters a field, the path that depended on it is known, not discovered by an incident.
How automation worksGovern
See change, dependencies and risk before they become incidents. Provenance and permissions travel with the understanding rather than being reapplied per integration.
Governance and protectionIntelligence
Understand the estate as a connected whole: where tools overlap, what is stored and never used, which fields another system could fill.
Stack IntelligenceMigrate
Plan change against evidence about real relationships. Know what breaks before it breaks, and preview the migration before a record moves.
MigrationAI
Give an agent grounded technology context: the schema, the fields, the relationships, the version, the provenance and the permissions, as structured context it can act on.
AI agents/ evidence
Built, not promised.
Every figure below is defined, dated, sourced and evidenced on the claims page.
As at 4 September 2026
As at 4 September 2026
As at 4 September 2026
As at 4 September 2026
As at 4 September 2026
As at 7 September 2026
What the numbers mean.
Precision matters more than scale, so the terms are defined.
A figure is only evidence if you know what it counts. Two of these are routinely confused, and the confusion inflates the impressive one: a vendor is an organisation, and a vendor–product API is one product from one vendor. Counting the second and calling it the first is how a corpus appears several times larger than it is.
| Term | Definition |
|---|---|
| Vendor | An organisation that publishes one or more APIs. |
| Vendor catalogued | A vendor indexed in the catalogue. Catalogued is not specced — a catalogued vendor may have had none of its APIs read. |
| Vendor–product API specced | One vendor×product grain: a distinct product from a distinct vendor, read to full specification. Not a count of organisations. |
| System | A distinct product or service exposing an interface. |
| API spec | A machine-readable description of an interface, in any of the schema formats Inferrex parses. |
| Field | A single named element within a schema, read and recorded. |
| Canonical key | A concept in the shared model that fields from different systems resolve to. |
| Integration pair | Two systems whose fields have been related through the canonical layer. Derived, not hand-built. |
Every figure on this site carries a definition, a measurement date, a method and a review date. Read the claims.
/ see it
Look at the understanding itself.
The Corpus
Every vendor Inferrex has catalogued and every API it has read, searchable. Check a system you actually run against it.
Explore The CorpusThe Universe
Every system the platform understands, drawn as one map. Not a logo wall: a picture of accumulated understanding and the relationships inside it.
See the UniverseA live run
Watch a provider change break a path, and watch the path be re-derived from the understanding rather than repaired by hand.
See it work/ where to start
Start against a system you already run.
Comprehension is only interesting against an estate that exists. The Development environment is free, self-directed and documentation-led: no scoping call, no procurement step and no commercial conversation first. Point it at a real API and ask it a question you cannot answer today.
Development
Free. Self-directed, documentation-led, and the entry point to everything else.
Start in DevelopmentThe environments
What each environment is for, who it is for, and what it makes possible as the work gets more serious.
See the environmentsThe developer surfaces
InferrexAPI, InferrexSDK, InferrexCLI and InferrexMCP drive the same model programmatically.
Read the docsYour technology already contains the answers.
The difficulty is the relationships between systems, and that is a comprehension problem before it is an integration one.

