The ontology your
firm's AI runs on
Every company, fund, person and document in your firm's data, resolved to one record that keeps itself current, with a source behind every connection.
Request DemoAgents that know your firm
Every agent plans over your entities and your definitions, so your Q3 is its Q3.
Answers across connections
Questions that span companies, funds and people follow the graph, with a source at every hop.
Maintains itself
Drafted from your documents, current with every new one, and better with every fix.
Every name,
one record
A dozen spellings across your CRM, fund reports and inbox, resolved to one record with your fields on it.
Your data, resolved
Every company, fund, person, deal and document, one record each.
Your definitions
Metrics, calendars and peer sets, written once in code.
The public record, linked in
Filings, financials and relationships attach wherever an entity files.
Built on your data,
the way your firm sees it
We start from a template for your kind of firm and draft the rest from your documents in the first week, work that used to take a six-month consulting project. We built this layer before at Palantir.
Hedge funds
Notes, models, emails and filings, each resolved to its company, with your coverage and statuses.
Allocators
Funds, managers and portfolio companies, with commitments and look-through holdings by period.
Law firms
Matters, clients, documents and people, and the parties and agreements between them.
Manufacturers
Customers, contracts, products and plants, connected across every system.
Source: Stop Paying to Forget.
See how much to trust
every connection
Every relationship shows how it is known and how many sources back it.
How it's known
Said outright, inferred from what was said, or computed from other claims.
How many sources say it
Lines thicken as more independent sources assert the same edge.
Whether sources agree
- Sources agree
- Sources differ in scope
- Sources contradict
When sources disagree, we show the split instead of picking a side.
An ontology that
corrects itself
Agents re-check it, extend it and learn from your fixes. Evidence and code decide what enters.
Agents re-check every edge
As new documents arrive, agents re-read the sources behind each edge and repair what no longer holds.
Grows its own schema
When documents describe something it has no type for, an agent proposes one for your team to approve.
Every fix teaches it
A correction made once becomes a rule, applied everywhere the same mistake appears.
Every edge knows
when it was true
Each claim keeps two clocks: when it was true, and when we learned it.
When it was true
Every relationship carries the period it held.
When it was recorded
A superseding filing closes the old fact out instead of erasing it.
As of any date
See the graph as it stood on any day, with nothing from later leaking in.
The first public
equities ontology
We started with the public record: built claim by claim from SEC filings, earnings calls and regulatory dockets, and updated with every new filing.
Every edge is one filer's word,
with the sentence attached
Every claim keeps the verbatim sentence it came from. Click any relationship and you land on that sentence, not just the document.
“We recognized an aggregate of approximately 67% and 65% of our revenue for the three months ended September 30, 2025 and 2024, from our largest customer, Microsoft.”
From filings
to a graph you can audit
Read every source
Filings from the company and every filer that names it, plus earnings calls and dockets.
Extract claims, with their quotes
Each relationship comes out with its verbatim sentence, checked against the source.
Check which way it points
A customer is never recorded as a supplier, nor a parent as a subsidiary.
Resolve the entities
NVDA, NVIDIA and NVIDIA CORP become one node, matched to SEC records.
Who is connected to whom, on whose word, in which sentence
Agents walk the graph as they work, following every path between two entities with the evidence at every hop. Your own agents reach the same graph through the Kepler MCP and API.