Built to last
What your documents actually say, connected up.
Not something you author and maintain. It is derived from your content, it remembers which document every claim came from, and it can be thrown away and rebuilt without re-reading a thing.
What it does for you
One answer, drawn from wherever the facts actually live.
Every account, product and plan that turns up in more than one of your systems becomes one thing the graph can reach across. Here is what that looks like.
One question
Does this account's plan include the instructor-led lab?
The answer, with its receipts
The account is on the Professional plan1, which includes the certification track but not the instructor-led lab inside it2 — that lab is a Premium add-on, billed separately3.
- [1]Salesforce — Account record
- [2]Plan Entitlements Matrix
- [3]Course Catalogue 2026
- CRM
- Entitlements
- Catalogue
- Billing
One question, drawn from a real trace. On your own content it does this live — which is the only way to really feel it.
In short
As we read your content we note what it claims and which system said it. When the same account or product or plan turns up in a billing record and in a support ticket, those become one thing — which is what lets a single answer span two applications. Because the graph is derived rather than authored, losing it costs a rebuild rather than the work.
The primitives
One thing that matters, and the three that make it possible.
the reason any of this matters
A thing that appears in two of your systems
The account named in a billing record and in a support ticket. These crossings are what let one answer span two applications, and they are counted per entity — so you can see the join working rather than having to take our word for it.
one per real-world thing
The thing itself
That same account mentioned in a CRM row and in a ticket becomes one thing rather than two, which is what makes the crossing above possible — a fact attached to either mention is reachable from the other.
one per fact, not per mention
Something your content claims about it
A single clean statement, however many documents happen to say it. Five documents making the same claim do not become five facts; they become one fact with five documents behind it.
the receipts
Each document that claims it
One record per document that states the fact, each carrying its own confidence, none of them averaged into a single number. This is what makes multi-source evidence faithful rather than approximate.
Provenance
Every document behind a fact, kept separately.
This is the promise the rest of the product rests on, and it is worth being explicit about what it buys.
- Every document that asserts a fact keeps its own confidence for it. Nothing is averaged away.
- A fact is retracted only when its last supporting document stops asserting it — so removing one source out of four changes the count, not the graph.
- The documents behind every fact an answer used are known exactly, so you can open each of them rather than trusting whichever passage looked closest.
- Changing your vocabulary — and changing your mind again — rewrites nothing you sent us and re-runs no extraction.
Asserted by
- 0.95 confidence
Q3 Retention Review
notion
- 0.88 confidence
Support Themes 2026
zendesk
- 0.81 confidence
Activation Playbook v4
lms_content
Inspection
Four consoles, because a graph you can't look at is a graph you can't trust.
Graph visualizer
Force-directed, expand any node one hop at a time, filter by entity type, predicate or source. Derived edges are dashed; edges whose supporters span two integrations are coloured, because those are the interesting ones.
Entity quality
Your most connected entities, the orphans with no connections at all, and a paged review queue of possible duplicates for a human to confirm — because merging two things that were not the same is not something you can undo.
Corpus health
A score out of 100 over ontology stability, entity match rate, orphan share, confidence distribution, extraction success rate and source coverage — with the direction of health stated for each, so a number is never ambiguous.
Content status
Per document: which phase it is in, how long each phase took, and the failure reason with a retry if it did not make it.
Corpus health
Every metric states which direction is healthy.
A dashboard number with no stated direction is a number people learn to ignore.
| Metric | What it counts | Healthy direction |
|---|---|---|
| Ontology stability | New types discovered per 100 documents | Healthy = declining toward 0 |
| Entity match rate | Share of mentions resolving to an existing entity | Healthy = rising toward 90%+ |
| Orphan entities | Entities with no assertions attached | Healthy = small and shrinking |
| Corroborated share | Assertions with more than one supporting document | Higher is better |
| Cross-source links | Entities bridging two or more integrations | More = better integration value |
| Extraction success rate | Documents completing extraction without failure | Healthy = at or near 100% |
Foundations
Lose the index or the graph, and nothing of yours is lost.
The search index and the graph can both be thrown away and rebuilt from what you sent us, without re-reading a document and without a second pass over any model.
That is why the things that are expensive elsewhere are cheap here. Changing your vocabulary needs no re-extraction. Losing the index is an afternoon rather than an incident. And neither of those sends you a processing bill or asks you to re-send a single document.
It is also the answer to the question every technical buyer asks third: what happens to our content if we leave. Everything of yours that we hold, you sent us — there is no proprietary intermediate you would have to get back out.
FAQ
Frequently asked
Including the limitations, because they change how you read the numbers.
Start with the free tier. Ship the first answer today.
RAG is free, with no card and no sales call. Add the graph when your corpus earns it.
- no card required
- first workspace is free
- one POST to ingest