Syndicated facilities between challenger banks
Counterparty authority verified in milliseconds, instead of stalling for human confirmation.
Inter-Agent Protocol
The future of Autonomous Finance
IAP Exchange removes the risks preventing today’s financial institutions from harnessing the potential of agentic AI.
Eight cryptographic and policy checks run before anything settles. If one fails, the tunnel never opens and a named human is summoned.
- Eight checks
- Before the tunnel opens
- Milliseconds
- To settle inside mandate
- 100%
- Non-custodial by design
- Human-in-the-loop
- Outliers demand human approval
Our purposeWe don’t compete.
We don’t compete.
We connect.
IAP Exchange provides the trusted infrastructure that lets autonomous agents create value, freely and securely, across an open financial ecosystem.
- Negotiate
- Transact
- Settle
- Comply
- Monitor
- Connect
01The problem
The hurdle is not capability. It is accountability.
The models already work. Deployment stalls because no chief risk officer will sign off on an agent whose limits cannot be proved to a regulator — and the exposure lands on a named individual, not a department.
IAP makes every agent decision bounded, attributable and provable. That is the condition under which autonomous finance becomes approvable.
You wouldn’t run a stock exchange without a clearing house. You cannot run an AI economy without a trust layer.
02The trust layer
Six pillars of enforcement.
Each pillar removes one reason an institutional risk function would refuse to let an agent act on its own.
01
Non-custodial neutrality
IAP never holds assets and never makes the decision. No custody risk, no conflict of interest.
02
Automated compliance
Policy is checked at the moment of interaction, not reconstructed from logs afterwards.
03
Cryptographic privacy
Agents prove a decision matched the rulebook without revealing the rulebook or the pricing logic.
04
Hardware-enforced gateway
Risk limits sit in a secure chip. Outside mandate, the mTLS tunnel never opens.
05
Provenance watermarking
Every proposal carries a C2PA-signed record of the data behind it. Unattestable inputs are held.
06
Mirror Test determinism
A deterministic model replays the counterparty's intent. Divergence triggers a cooling-off period.
02.1The distinction
Prevention, not observability.
Everyone else: observability
Log the calls. Find the breach in the review, after the money moved.
IAP: prevention
Checked before the session exists. There is no breach to write up, because the trade never left the building.
02.2In plain terms
A digital parent for the bank's AI.
Strip the cryptography away and the job is simple to describe.
01
An invisible referee
It never tells the bank's AI how to think. It stops it doing anything illegal or reckless.
02
A boundary it can't move
The bank sets the rules. The agent cannot renegotiate its own limits.
03
A physical stop button
If the agent goes rogue, the network connection for that trade simply never opens.
04
An un-fakeable receipt
Every interaction is stamped: who authorised it, what it could do, what data it used.
03Where it applies
Three deployments already being asked for.
Overnight positions inside the window
Treasury agents settle inside mandate. Humans see exceptions, not every transaction.
Evidence for AI-driven AML decisions
A signed manifest per interaction — auditable without exposing the model behind it.
04Why us
Specified in the room with the people who carry the liability.
IAP's founder sits with 10 to 12 chairs, chief risk officers and chief compliance officers every week. The protocol was specified from what they said was missing, not designed in the abstract.
- 200+
- Executive GRC forums convened annually
- 7,000+
- Senior financial services following
- 30 yrs
- Institutional risk experience on the board
05 / The ask
Request the data roomWe are raising a pre-seed round to take IAP from alpha to regulated deployment.
SEIS qualifying for UK investors. Terms, technical specification and delivery roadmap available on request.