Extraction that returns
your existing schema.
You already have a downstream pipeline. We power the extraction underneath and return the result in your JSON shape — not ours. Zero refactor on your side, ready in 48 hours.
First partner integrated 2026-05-22 — 9.2 seconds end-to-end, single attempt, 100% confidence.
The problem
Your pipeline is built around a schema. Ours isn't the same.
Every B2B integration runs into it. Your team built downstream code, workflows, dashboards, automation logic — all expecting your existing extraction vendor's shape. Switching providers used to mean refactoring all of it. Or: your engineering team writes a translation layer, and now every new provider is a coupling you maintain.
The usual answer
Refactor your code
Months of engineering work, every existing integration broken until the migration ships.
The DIY answer
Build a translation layer
Your engineers write per-vendor adapters and maintain them forever. Every vendor change breaks it.
The Adapter answer
We translate on our side
Send us your schema. We author a profile. Your pipeline never knows we're the engine underneath.
How it works
One endpoint. Per-partner shape profile. Zero code per partner.
You send us a document
POST a pre-signed HTTPS URL to your S3 bucket (or any storage you control) to /api/v1/partner/extract. We do a plain HTTPS GET — no cross-account IAM grants, no role coupling, bounded by your URL's TTL.
NexosIQ extracts
Same multi-pass extraction pipeline that powers our in-product invoice scans. Single-attempt extraction in under 10 seconds for typical documents, with the full retry + OCR fallback chain for harder ones.
Your shape profile transforms the result
A small declarative profile (per partner) defines the renames, nested groupings, computed values, and synthetic fields needed to turn our canonical extraction output into your existing JSON schema.
You get YOUR shape back
Drop-in replacement for your existing extraction provider. 24-hour idempotency on a correlation id — retries return the cached result with a cache_hit flag so your queue workers can't double-bill or duplicate downstream records.
For your engineering team
The pattern underneath: Anti-Corruption Layer.
Coined by Eric Evans in Domain-Driven Design (2003). A translation layer between a system's canonical model and an outside consumer's model, so neither side has to compromise. The same pattern Stripe Connect uses for per-platform destination payloads, Twilio uses for per-app webhook templates, and Segment uses for its 300+ downstream destinations.
We applied it to extraction: one endpoint, one canonical output, N declarative partner profiles. The profile is JSONB on a partner row — operationally a config change, not a code change. Profile updates take effect on the partner's next API call. No deploys, no migrations, no engineer-hours per partner.
POST /api/v1/partner/extract
Authorization: Bearer pnr_live_<your-key>
Content-Type: application/json
{
"source_url": "https://your-bucket.s3.us-east-1.amazonaws.com/...",
"correlation_id": "<your-ingestion-uuid>",
"subject": "Invoice #INV-12345"
}
→ 200 OK { ...response in YOUR shape, governed by your profile }Security
Built for the parts your security team will actually ask about.
You keep your data
Pre-signed URL pattern means we never re-host your files. You control storage, retention, and access on your side. URL TTL caps the exposure window.
Hardened HTTPS fetch
SSRF defenses: host allowlist, private-IP block (incl. cloud metadata services), 10s timeout, 50MB body cap, redirect refusal, URL userinfo refusal.
OAuth tokens encrypted at rest
Every credential we store — partner API keys, OAuth refresh tokens — is AES-256-GCM encrypted at rest with envelope keys we rotate ourselves.
Auditable per-call
Every extraction call lands in a tamper-evident audit log with partner id, correlation id, response time, and confidence — joinable to your own audit trail.
Full sub-processor disclosure at /sub-processors. DPA available on request before signing.
Pricing
Three levers. No surprises.
Adapter is priced separately from the NEXOS dashboard product — you can be on one, both, or just this. The fit fee covers the work we do to author and test your shape profile against your real payloads; everything after is pure consumption.
One-time
Adapter Setup
$5,000
We author your shape profile, integration-test it against your real payloads, and ship the working API key within 48 hours.
Per extraction
Usage
From $0.50
Volume-tiered. Drops as you scale — talk to us if you're planning above 50k extractions/month.
Monthly minimum
Floor
$500/mo
A baseline that keeps the infrastructure warm even on quiet months. Counts toward your usage.
Send us your API spec.
Email partners@nexosscan.com with your existing extraction response schema (or a link to your docs). We'll send back a sample profile authored against your shape + a 48-hour SOW. No demo call required.