Which one is the product?
Software claims accumulate across layers — a tracker ticket, a docs page, the actual code, a deployed release — and those layers drift out of sync silently. Horologium reconciles them into one queryable Product Truth, with cited evidence for every fact, instead of letting each layer assert its own version of reality.
A concrete conflict, not a hypothetical one
This is the exact scenario running in the live demo right now — a refund-window capability
(SHOP-412) where every source of truth disagrees.
REFUND_WINDOW_DAYS = 30, agreeing with Jira, not docs.v2.0.0 and is live on beta —
production has not received it.
documentation_conflict (docs vs. Jira/code) and a released_not_live
(Done in the tracker is not the same fact as live for customers). See it for yourself in the
live demo.
The reality chain Horologium reconciles
A product's reality is not one status field — it's a sequence of layers, each of which can be true independently of the others.
In this Public Alpha, Implemented, Released and Deployed are a real,
deterministic lifecycle vector computed per capability, per deployment target — never collapsed
into one badge. Enabled and Observed have no provider yet and
honestly resolve to Unresolved rather than a guess; Tested and
Built are modeled but not yet independent predicates. See the
full reality-layers explanation
for exactly what's solid today and what's a named gap.
Not another code map, docs generator, or knowledge graph
- Not a code map. Horologium doesn't just render your repository's structure — it cross-references code against Jira, docs, and deployment records, and tells you where they disagree.
- Not a docs generator. Horologium doesn't write documentation for you. It cites the docs you already have as evidence and flags when they contradict the tracker or the code.
- Not a generic knowledge graph. Every node and edge resolves to one of a fixed, closed set of Product Truth entities and relationships — never an open-ended, LLM-invented schema. Deterministic facts are computed deterministically; a model is used only where a human would also have to read prose and judge, and only to produce a proposal that's validated before it counts.
See it, run it, or bring us a real conflict
Live demo
The exact scenario above, running against a synthetic fixture. No signup, no install.
Open the live demoDocs
The full Product Truth model, the evidence and integrity model, and Public Alpha limits.
Read the docsSelf-host
Run Horologium against your own Git, Jira, and docs — self-host first, no cloud dependency.
Self-host quickstartMCP / AI agent
Let a coding agent query Product Truth directly — read-only, one tool per API endpoint.
MCP integration guideRequest a pilot
Bring a real conflict from your own product for a Product Integrity Audit conversation.
Early access — reviewed individually, no self-serve signup Request a pilot