Inferrex

/ submit an api

Know a system we should understand?

The Corpus grows two ways: our crawlers find what's published, and people who know a system tell us about it. Point us at an API — yours, your vendor's, the obscure one only your industry uses — and Inferrex comprehends it for everyone.

/ the pipeline

Submission → comprehension → review → release.

Nothing you submit publishes unreviewed, and everything that publishes carries its provenance.

Every submission enters the same pipeline our own crawlers feed. Inferrex reads what you point it at — a documentation URL, an OpenAPI/GraphQL/SOAP/whatever spec, or a manifest you've already shaped — infers the structure, classifies every field by meaning, and stages the result. A human reviews it before it becomes part of the library, and the published connector records how it entered the corpus and who improved it.

Anything with a machine-readable surface

A docs URL is enough to start; a spec is better. Every major protocol counts — REST, GraphQL, SOAP, OData, gRPC, EDI, even COBOL copybooks.

Structure only, never data

A contribution holds field names, types and endpoints — the shape of a system. No records, no credentials, no customer data, ever.

Reviewed, not auto-published

Submissions land in a moderation queue and ship only in a versioned release an admin has signed off. The corpus stays curated.

Provenance on the tin

Published connectors carry their source — crawled, documented, or improved by a named contributor. Credit where the understanding came from.

/ the compounding part

One good submission helps every future integration.

Because every system in the corpus speaks the same canonical language, your contribution doesn't just add one connector — it gives every other system a new counterpart to map against.

If you run a product with an API, being comprehended is distribution: every Inferrex customer can connect to you without you building or maintaining anything. If you depend on a niche system — the industry-specific ERP, the regional payroll provider, the legacy platform nobody wrote a connector for — submitting it once means you, and everyone after you, stop paying the no-connector tax.

And if a connector already exists but is thin — entities missing, fields untyped — the same flow accepts improvements. The directory marks what needs work; the corpus page shows the live state of every vendor.

Point us at a system.

A URL is enough to start. The comprehension is our job.