
Aggregate insights for consented cohorts. Personal enrichment for a customer you already have. Parameters choose the memory layer — not a menu of extra products. Identity fields are never in the payload.
Purpose, scope, user control, then a receipt. Commercial terms come after we review your use case.
1
Aggregate insights for a consented cohort, or personal enrichment for a customer you already have.
POST /api/marketplace/aggregate-query
POST /api/marketplace/personalized-query2
Each product is one API. data_layer, mode, and fidelity choose the slice — not a fourth product.
3
Aggregate uses listing enrollment. Enrichment: they approve your brand and can turn quoted snippets off per layer. Name, email, phone, and address are never in the payload.
4
Users can inspect what was shared in the Audit Log and revoke going forward.
Same memory graph. Two ways to query it.
Cohort patterns. No customer relationship required.
POST /api/marketplace/aggregate-querymode: fast | premium
data_layer: knowledge_graph | behavioral_signals | temporal_facts (premium)
max_providers: 1–1000 (premium)
1
Query enrolled users as a group.
2
Receive distributions and top lists — never a dossier you can match to a named person.
3
Graph attributes only. Evidence snippets are not available at cohort level.
Attributes for someone you already have, after they approve you.
POST /api/marketplace/personalized-querytarget_user_id, query
data_layer: knowledge_graph | behavioral_signals | temporal_facts
fidelity: graph | evidence
1
Start from a customer already in your systems.
2
You request ingest origins (email, calendar, …). They approve or reject, and can turn quoted snippets on or off per layer — they do not pick origins.
3
graph returns anonymized attributes from those origins. evidence adds quoted retrieve snippets — not full artifacts, open URLs, identity fields, bank feeds, or device companion data.
One sample per parameter path. Commercial terms are provided after review.
Sample (no PII) — Aggregate insights · mode=fast
POST /api/marketplace/aggregate-query
{
"success": true,
"product": "aggregate_insights",
"mode": "fast",
"totalProviders": 142,
"insightsProviders": 142,
"personalizedProviders": 28,
"entityTypeDistribution": {
"person": 120,
"organization": 85,
"location": 64
},
"signalCategoryDistribution": {
"preference": 95,
"routine": 70,
"interest": 62
},
"factPredicateDistribution": {
"prefers": 72,
"attended": 39,
"works_at": 31
}
}Every assistant is building a private profile of its users. That profile is stuck where it was made — ChatGPT memory does not follow someone to Claude, Cursor, or your product. Cognitia is the user-owned graph they can inspect, and the consent surface they use to share a slice with you.