Your knowledge layer

Your research, structured once, owned forever.

Your research is your alpha. Structure it. Own it. Compound it.

Your research, structured once at ingestion and owned in your tenant. Three modules ship as one layer: Trade Idea Hub, Narrative Network and Voice Scoring.

Request a demo

Why it matters

Today it is used once and lost: pulled in as context, consumed by one answer, and often processed outside your walls.

Your knowledge layer inverts this

Four claims

Your research is your alpha.

Own the corpus

Structured once at ingestion, deployed in your tenant.

Capability without exposure

Frontier models by default, private or open-source by choice.

Compound, don't converge

Sharper with every document, not averaging toward the street.

Agents by choice

Consume it headless, build on it with AgentOS, or use Sonαr out of the box.

Follow one document through
The knowledge layer · How it works

Structured once. Used everywhere.

Research flows in, is structured once in your knowledge layer, and powers every agent, desk and pipeline downstream. Chat tools answer and forget. This layer remembers.

Follow one document through Illustrative examples
01Sources

Internal and external research, filings, market data

02Processing

Parse, classify, resolve, extract, review

03Knowledge layer

Indices, typed records and a graph of attributed beliefs

04Surfaces

Sonαr, your agents, BI and quant

Internal Research

Analyst notes, recommendations, RMS

PM note
IC memo
Analyst rec

External Research

Broker and boutique research

Report
Morning
Chat

Company Intelligence

Fundamentals, filings, transcripts

Transcript10-K8-K20-F

Reference & Market Data

Entity master, prices, ratings, holdings

CDX.NA.IGcredit · index
UST 10Ygovt · rates
USD/JPYFX · spot
Ingestion
Microsoft Suite
Email & SFTP
Direct Channels
Vendor APIs / Webhooks
MCP Inbound
Processingevery field reviewed

  1. 1Multi-modal parsing
  2. 2Domain classification
  3. 3Entity & instrument resolution
  4. 4Multi-agent extraction, reviewed
  5. 5Taxonomy normalization
  6. 6Temporal versioning
Your knowledge layeryou own the corpus · in your tenant
Knowledge GraphNarrative Network

voices on this edge
driverasset
Structured Data StoreTrade Idea Hubtyped records
Semantic IndicesFoundationretrieval by meaning
Orchestrator · one entry point across every store · lineage intact
Egress
Retrieval API
MCP Server
Structured Egress
Quant Data Egress

Sonαr

Reasoning, agents, Voice Scoring

What did hyperscalers say on capex?

Raised again. 9 management voices tie AI demand to build-out.call+8

Your Agents

Internal copilots, PM-facing apps

› narrative_alignment(book="rates")
← UST 10Y: same call, other reason
← street: the Fed · you: issuance

Warehouse / BI

Analytics, risk, downstream pipelines

instrsideentrytargetstop USD/JPYshort158.50152.00161.00 .....

Analysts & Quant

Signals, derived facts, time-series

voices added per month · sustained

Now

PersistsEvery claim becomes a structured, attributed edge on a canonical entity. Stored, not re-derived.
CompoundsEach document adds voices, breadth and momentum to one shared graph.
EvolvesEvery claim is dated as made and as received. Consensus and reversals are never backfilled.
AttributedA dated map of what the market believes, traceable to the sentence. Not a claim that those beliefs are true.
The knowledge layer · How it workslogo

Structured once. Used everywhere.

Research flows in, is structured once in your knowledge layer, and powers every agent, desk and pipeline downstream. Chat tools answer and forget. This layer remembers.

Follow one document throughIllustrative
Sourcesresearch, filings, market data

Internal Research

Analyst notes, recommendations, RMS

PM note
IC memo
Analyst rec

External Research

Broker and boutique research

Report
Morning
Chat

Company Intelligence

Fundamentals, filings, transcripts

Transcript10-K8-K20-F

Reference & Market Data

Entity master, prices, ratings, holdings

CDX.NA.IGcredit
UST 10Yrates
USD/JPYFX
Ingestion
Microsoft SuiteEmail & SFTPDirect ChannelsVendor APIs / WebhooksMCP Inbound
Processingevery field reviewed

  1. 1Multi-modal parsing
  2. 2Domain classification
  3. 3Entity & instrument resolution
  4. 4Multi-agent extraction, reviewed
  5. 5Taxonomy normalization
  6. 6Temporal versioning
Your knowledge layerin your tenant
Knowledge GraphNarrative Network

voices

Data StoreTrade Idea Hub
Semantic Indices
you own the corpusillustrative graph
SurfacesSonαr, agents, BI, quant
OrchestratorOne entry point across every store, lineage intact
Egress
Retrieval APIMCP ServerStructured EgressQuant Data Egress

PersistsEvery claim becomes a structured, attributed edge on a canonical entity. Stored, not re-derived.
CompoundsEach document adds voices, breadth and momentum to one shared graph.
EvolvesEvery claim is dated as made and as received. Consensus and reversals are never backfilled.
AttributedA dated map of what the market believes, traceable to the sentence. Not a claim that those beliefs are true.
Three modules, one layer

Three modules

Three modules ship as one layer.

Structured extraction

Trade Idea Hub

A single, structured idea hub and directory, with full lineage.

Knowledge graph

Narrative Network

A dated, attributed graph of what the market believes.

Credibility signals

Voice Scoring

An auditable record of who was right, when, and under what conditions.

How you consume it

Three ways in

One layer. Three ways in.

Infrastructure first, applications by choice. You own the corpus, deployed in your tenant. How you consume it, and scale it, is your decision.

Delivered where work happens

EmailTeams / SlackOfficeSnowflakeMCPAPI

Headless

Your applications & agents

Output-only delivery of structured ideas, signals and scores into your warehouse and your own enterprise agents, delivered as features into nowcast pipelines and quant models.

Build on it

AgentOS

Agent and LLM infrastructure: orchestration, memory, tool integration, model routing and guardrails, to build your own solutions across the value chain.

Out of the box

Sonαr

A virtual analyst built on AgentOS and tuned for institutional multi-asset investing. Not an aggregator: it reads, reasons and remembers.

Sonαr StudioSkills layer

Your knowledge layerOne corpus, structured once, serving all three paths. Rates, credit, FX, commodities and equities. Macro and fixed income first.

Trade Idea HubNarrative NetworkVoice Scoring

FAQ

Questions institutional teams ask.

What is a knowledge layer?

Your accumulated research, notes, trade ideas and conviction history, structured once at ingestion and deployed in your tenant, so it compounds instead of being used once and lost.

Who owns the knowledge layer?

You do, contractually and architecturally: the corpus, the derived structures and your desk-authored Skills, in your environment and on exit.

How do teams consume the knowledge layer?

Three ways: headless into your pipelines and models, through AgentOS to build your own solutions, or out of the box in Sonαr.

All questions

Get started

From a working session to a
production-grade proof.

Ninety minutes with your research leads and PMs to define the use cases, the corpus in scope, and what success has to look like in numbers.

Request a demoExplore Sonαr
Built for
Global asset managers and hedge funds
Deployment
Single-tenant, in your cloud boundary
Models
Frontier by default, private by choice
Compliance
SOC 2 Type II · GDPR · ISO 27001 (in progress)