Platform
One ledger.
Many products you could build.
What ships today is deliberately small: the ingestion API, the computed layer, and a dashboard. What it makes buildable is the point — every capability below is a query against statements you already have, not a new pipeline.
Shipped today: the API and the dashboard. Everything below is what the substrate makes buildable — not products we ship.
Agent observability
Every tool call, decision, and outcome as a statement — then ask what an agent did last month, not just last session. Lifetimes, not sessions.
Skill graphs & proficiency
Belief state (FSRS) is computed per actor and concept from real behavior — a provable record of what someone knows, decaying honestly over time.
Difficulty & struggle maps
Per-item difficulty computed from the whole population's traces. Find where users struggle without instrumenting anything twice.
Audit & compliance evidence
The immutable ledger plus full export is the raw material for GDPR / EU AI Act workflows — evidence by architecture.
Personalization & reranking
The rerank endpoint orders candidates by an actor's actual history — recommendation infrastructure without shipping your data to a black box.
Multi-actor analytics
Humans, agents, devices, and systems in one queryable model — cross-actor questions become one query instead of four ETL pipelines.
Proven how?
Empress runs in production as the substrate behind Swapp, a consumer feed of interactive cards — hundreds of thousands of real statements, per-interaction traces, ELO, belief, and personalization all computed from the ledger. The first tenant is the proof; the second key is yours to request.