Monetary Alignment

Governed Agentic Intelligence
for Interbank Markets.

An embedded intelligence layer operating within the
governed IDBX execution environment.

AIDANN is being built to operate within the IDBX execution and settlement architecture, supporting visibility and analysis while remaining bound by venue rules and institutional guardrails.

Request an AIDANN briefing
Explore use cases

Governed intelligence

Inside the venue

Analytics, navigation and netting.

AI for Data-driven Analytics, Navigation and Netting

An intelligence layer inside the IDBX venue, not a terminal beside it. It reads the book, finds the route, and nets what can be netted, within venue logic it cannot override.

AIDANNSees everything. Trades nothing.
Scroll

London New York Singapore

Live · computed in your browser

The engine below is running now. Obligations arrive, it recomputes, and AIDANN describes what changed. Watch the gap between gross and net.

Live Demonstration
AIDANN netting, computed live.
THE ENGINE COMPUTES · AIDANN DESCRIBES · NOTHING IS GENERATED
ENGINE RUNNING
What you are watching. A deterministic netting engine runs in this page. Obligations arrive, the engine recomputes, and AIDANN narrates the result. Every figure below is computed. AIDANN's sentence is assembled from those computed values using a fixed template, which you can inspect. No language model is involved, and none is needed.

Obligation Feed · 0 trades

Audit, rechecked every tick

Five independent checks recompute from the raw obligations on every recalculation. A netting engine that cannot be audited is not a netting engine.

Engine State

Gross Obligations00 instructions
After Bilateral00 instructions
After Multilateral00 instructions
Liquidity Released00.00% of gross

Net Position Per Participant

Cycles Cancelled · 0

A directed cycle is a chain of debts returning to its origin. The minimum edge can be extinguished from every leg without changing any participant's net position.
AIDANN

Waiting for obligations.

Deterministic netting engine · integer cent arithmetic, no floating point · instruments: real sovereign ISINs; prices and trades illustrative.
No language model produced, estimated or rounded any figure in this demonstration.

What you just watched

Two passes, then the funding number falls.

01

Bilateral netting

Two participants who owe each other in the same currency and settlement cycle do not need to move both amounts. The engine offsets the pair and keeps the difference, in one direction. Nothing is estimated: obligations are summed per ordered pair, the smaller side is extinguished, and what remains is a single instruction between those two names.

This is the pass most desks already run, usually per counterparty, often in a spreadsheet, and usually once a day.

02

Multilateral cycle cancellation

The second pass is the one that is rarely run by hand. After bilateral netting, obligations form a directed graph, and that graph contains cycles: A owes B, B owes C, C owes A. Every leg of a cycle can be reduced by the size of its smallest leg, and no participant's net position changes by a single cent: each one pays that much less and receives that much less.

The smallest leg is extinguished entirely, so the cycle disappears and one instruction goes with it. The engine repeats this until no cycle remains. The drawing below runs the whole operation, one step at a time.

03

What falls out

Two numbers move, and they matter to different people. Funding falls, because a participant only needs cash for its net position rather than its gross obligations, that is intraday liquidity released back to the treasurer. Instruction count falls, because each extinguished leg is a settlement message that never has to be sent, matched, chased or failed, that is operations and settlement risk.

The engine above reports both, every tick, alongside five checks that recompute from the raw obligations.

BEFORE · THREE OBLIGATIONS A B C 70 50 30 NET · A ‒20 · B +20 · C 0 AFTER · THE SMALLEST LEG IS GONE A B C 70 50 30 NET · A ‒20 · B +20 · C 0

Every leg drops by the size of the smallest. A is still 20 down, B still 20 up, C still flat, but 30 never leaves anyone's account, and the obligation that carried it no longer exists.

Two things netting does not do. It does not change economic exposure: a position is the same position before and after, and market risk is untouched. And it is only as good as the agreements behind it, settlement netting depends on enforceable arrangements between the parties and on the law of the relevant jurisdiction. The arithmetic above is not in question; the legal basis is a matter of documentation, not computation.

Data-driven Analytics

Analytics here means arithmetic on the book.

Not a forecast, and not a score. AIDANN takes live trade data from the participating banks and every action on the platform, builds the consolidated book from it, and computes what is in it: where the touch is, what the spread costs, how much size sits inside a band of the mid, and which way the weight is leaning. Every figure below is recomputed from the ladder on the left as it changes, and every one carries the formula that produced it.

Consolidated book · illustrative · size in €m

100.000mid

The ladder is an illustrative book and is labelled as one. The arithmetic performed on it is real, deterministic and reproducible by hand: take the figures, apply the formula printed beside each readout, and you will get the same answer. Nothing here is estimated and nothing is predicted.

Navigation

Navigation is showing where the size is.

A clip larger than the touch has to come from somewhere. It can be filled at one clearing price for the whole amount, or it can walk the levels and pay each one its own price. The difference is not a matter of opinion, it is subtraction. Both are set out below, computed against the same book. Neither is recommended: which one to use, and whether to trade at all, is the trader’s call.

Clip

Route A · one price for the whole clip

Route B · walk the levels

AIDANN sets out both routes and the arithmetic between them. It does not prefer one, it does not suggest trading, and it cannot send either. That is a person, or a venue rule.

Both totals are consideration on the same illustrative book: size multiplied by price, divided by 100, and printed to three decimals so the difference between the two routes is visible rather than rounded away. Route A pays the worst price it has to touch on the entire clip. Route B pays each level its own price.

The trade assistant

A trader asks. It answers from everything it can see.

This is the other half of navigation, and the plainer half: moving around the venue with something that has the whole picture and will tell you what is in it. Bank feeds, platform activity, the book, obligations and settlement state, all of it available to the question in front of you. It answers what it is asked, in figures, and then it stops. Press a question and watch the reply computed from the same book above.

AIDANN

Sees everything. Trades nothing. AIDANN will not tell a trader to trade, will not recommend one route over another, and cannot place, amend or cancel an order or alter a settlement instruction. It answers the question it was asked. The decision, and the trade, belong to the trader.

Each reply is assembled from a fixed template with computed slots, on the illustrative book above. There is no language model here: ask the same question twice on the same book and you will get the same sentence, with the same figures, and every figure can be checked against the ladder.

The record

Everything it captures becomes the banks’ own record.

AIDANN captures the live trade data the banks send it and every action on the platform. Held and ordered over time, that stops being a feed and becomes a history of the venue: what was quoted, what traded, what size was there, what netted against what, and when. IDBX redistributes that record back to the participating banks. It is their own activity, returned to them.

What accumulates

  • Every trade received from the participating banks.
  • Every action on the platform, timestamped and attributable.
  • The book as it stood at the moment of each of them.
  • Obligations, and what netted against what.

What it becomes

  • A continuous history of the venue, not a sampled snapshot.
  • Prices and size with the context they occurred in.
  • Netting outcomes, cycle by cycle, over time.
  • A record that is longer and denser every day it runs.

Where it goes

  • Back to the participating banks.
  • Their own activity, returned rather than sold on.
  • Market data is a cost every institution already carries.
  • Here the venue gives the record back to the desks that made it.
THE BANKS PLATFORM ACTIVITY AIDANN HISTORICAL RECORD REDISTRIBUTED TO THE BANKS THAT MADE IT
The banks send it, the venue records it, and the record goes home. The longer it runs, the more it is worth to the institutions that built it.

This band describes what the venue does with the data it holds. It states no figure for what that is worth, because a saving depends on what an institution pays today for comparable history, and that is a number each participant knows better than we do.

Integrated by Design

Intelligence as Infrastructure.

AIDANN is embedded within the IDBX execution and settlement stack, operating inside the
same deterministic frameworks that govern matching, coordination and settlement. It does
not sit above the system. It operates within it. Every output remains constrained by venue
logic, institutional governance and regulatory guardrails.

Request an AIDANN Briefing
View Governance Framework

Governance & Containment

Intelligence Without
Autonomy.

AIDANN does not override
execution rules

It does not bypass
settlement controls

It does not operate outside
institutional parameters

AIDANN will operate within deterministic venue logic and auditable governance frameworks. It holds no execution permissions and cannot place, amend or cancel an order.

Governed Agentic Intelligence
for Interbank Markets.

An embedded intelligence layer operating within the
governed IDBX execution environment.

Governance Participation

Founding institutions contribute to rulebook formation, rollout sequencing and advisory frameworks shaping the rail’s evolution.

Structured Economic Alignment

Participation includes defined economic alignment during the early implementation phase, linking execution efficiency with long term structural value.

Institutional Intelligence Priority

Priority access to enriched, anonymised market intelligence generated within IDBX and AIDANN, governed by strict data and oversight frameworks.

Settlement Migration Positioning

A front row position as CeBM integration and atomic settlement corridors transition from phased implementation to standardised market practice.

Auditability

Fully Traceable by Design.

AIDANN is being built so that every output is logged, attributable and reviewable within institutional governance frameworks.

Trust & Containment

Zero Signal Leakage.

AIDANN is designed so that trade intent, analysis and interaction remain fully contained within IDBX infrastructure, operating within defined infrastructure boundaries.

IDBX is not a crypto exchange.

Not a retail platform.

Not a data terminal.

Constrained by Rules. Aligned with Institutions.

Guardrails

AIDANN operates within defined operational and regulatory boundaries.

What AIDANN Is Not

Not autonomous.

Does not execute or override policy.

Does not replace trader discretion.

Not a public or general purpose chatbot.

What AIDANN Is

Consolidated pricing and liquidity visibility in a single governed view. Automated eligibility, credit and policy controls enforced prior to order submission.

A governed intelligence layer embedded within the IDBX rail, augmenting institutional decision making while remaining constrained by venue rules and regulatory guardrails.

Why It Matters
The engine is arithmetic. That is the point.

Given a set of obligations, the netting engine builds a directed graph, finds the cycles, cancels the maximum flow around each, and computes every participant’s net position against the system. The same inputs always produce the same output. It can be handed to a regulator and verified by hand.

OBLIGATIONSthe raw trades
→
NETTING ENGINEdeterministic code
→
RESULTaudited figures
→
AIDANNexplains only

The model never computes. If AIDANN were switched off tomorrow, the netting would still run; participants would read a table instead of a paragraph. This is what makes the output defensible to a model risk function: it is not a model output. It is a computation, described.

Netting compounds with membership

Netting efficiency is not a property of the venue. It is a property of the flow through it. Across thousands of randomised obligation sets, the reduction in gross settlement obligations rises with the density of trading between participants. A thin venue nets poorly. Each participant that joins improves the netting for every participant already there.

3 banks10 trades
66%
3 banks150 trades
92%
4 banks25 trades
77%
6 banks60 trades
81%
10 banks10 trades
40%
10 banks150 trades
84%
20 banks10 trades
23%
20 banks150 trades
77%
Median gross reduction across 120 randomised obligation sets per configuration. Deterministic seed. Illustrative simulation.
This is why the demonstration above can be trusted. Every figure it shows is computed, audited on every tick, and merely described by AIDANN — never generated.

Institutional access

You have watched the arithmetic.
The next conversation is about your desks.

Onboarding starts with a technical and commercial review: your flow, your counterparties, your settlement cycle, and what netting and routing would return on them. Founding participation is agreed institution by institution.

Request onboarding → Speak to the team first →

IDBX is a venue for regulated institutions. Access is granted to eligible participants under the venue’s onboarding requirements.