Your methodology,
first-class.
Every firm has a proprietary view of risk — today it lives in spreadsheets and one analyst’s Python scripts. Custom Axes turns that methodology into a first-class citizen of the ontology: queryable, alertable, reproducible. Your IP stays yours; the agent runs in your environment.
- name
- credit_conviction
- range
- −1.0 … +1.0
- cadence
- every 6h
- calibration corpus
- 4,200 labeled obligors
Define the shape; DEINO handles versioning, manifests, and reproduction for free.
Preview · switch methodology and walk Define → Agent → Calibrate → Deploy.
The methodology you can’t buy off the shelf.
Your credit team scores counterparties its own way. Your AML team has its own behavioral profile. Today that edge is tribal knowledge — one departure from being lost. Custom Axes makes it durable infrastructure.
Define the axis
Name, range (−1…1, 0…100, categorical), update cadence, and the calibration corpus to grade against.
Write the agent
A Python, TypeScript, or pure-prompt function. Input: NodeVersion context. Output: value + confidence + sources.
Calibrate
DEINO runs your agent against the corpus and reports precision, recall, and calibration error. Below the floor, it won’t ship.
Deploy
The axis goes live — queryable in Living Map, alertable via Drift Alerts, auditable in Forensics. Like any built-in axis.
Marketplace (V3)
Publish to the DEINO marketplace. Revenue share with other regulated firms who license your axis.
Your IP stays yours
The agent runs in your environment, with your credentials. DEINO never sees the methodology in cleartext.
An agent is a function.
Context in, calibrated value out. The same shape the studio previews — registered once, versioned and manifest-signed forever after.
from deino import Client, NodeContext, AxisValue client = Client(api_key="dno_live_...") def profile(ctx: NodeContext) -> AxisValue: f = ctx.attributes("leverage", "coverage") score = our_credit_model(f) # your IP, your env return AxisValue(value=score, confidence=model.conf(), sources=ctx.citations()) client.axes.register( name="credit_conviction", agent=profile, corpus="obligors_4200.parquet", ) # calibrated, versioned, queryable — like any built-in axis
Define once, live everywhere.
A Custom Axis isn’t a silo — the moment it deploys, it behaves like a native axis across every product on the ontology.
What are Custom Axes?
A way to plug in your own profiling agent and turn your firm’s proprietary methodology into a first-class axis of the ontology — queryable in Living Map, alertable via Drift Alerts, and reproducible in Forensics, like any built-in axis.
Does DEINO see my methodology?
No. The agent runs in your own environment with your credentials, and DEINO never sees the methodology in cleartext. Your IP stays yours.
How does an axis get approved before going live?
DEINO runs your agent against a calibration corpus and reports precision, recall, and calibration error. If it falls below the floor, the axis won’t ship — so only calibrated axes go live.
In alpha with three LATAM Tier-2 banks. The marketplace launches with V3 in Q4 2027; implementations are founder-led through alpha and beta. Request a slot →
Codify your edge.
Founder-led conversations for alpha partners — limited slots. A free tier is active today for the V1 products Custom Axes builds on.