Agent connectivity
A shared interface between AI channels and institution workflows.
Platform directionAI understands the intent.
CADON connects it to a controlled financial action.
We’re building the infrastructure between AI assistants and financial institutions — with sensitive customer workflows outside the model and the institution in control.
The connected assistant checks safe context, then CADON displays its provider cards.
prepare_cadon_contextthen, when readyshow_car_loan_bank_optionsThis conversation example illustrates the actual plugin’s tool sequence. No application form is collected in chat.
An AI can understand a financial request. Turning it into an authorized action needs a workflow the institution can control.
Identify the relevant financial product or workflow.
Create a session for the selected institution.
The user verifies identity, reviews details and gives consent.
The institution processes the authorized request.
Send the necessary outcome back to the assistant.
Target execution lifecycle. The connected MCP demo uses mock providers and does not execute live financial actions.
The assistant needs the intent and an appropriate status. Identity documents, account credentials and application details belong in controlled financial systems.
Only the context needed to initiate and follow the workflow.
Collected through CADON’s intended secure session or the institution’s systems.
Financial institutions define which products and actions are available. AI developers get a consistent way to initiate them.
CADON’s platform direction is to connect both sides through shared execution infrastructure, rather than rebuild each connection.
Explore an integrationPlatform direction — these are ecosystem examples, not partner logos or live integrations.
Our prototype makes the handoff tangible. The broader platform is being developed around these capabilities.
A shared interface between AI channels and institution workflows.
Platform directionA dedicated environment for sensitive financial interactions.
Platform directionThe institution defines actions, limits and required approvals.
Platform directionExplicit authorization and traceable events across the workflow.
Platform directionTalk to the live assistant on our website, or use CADON inside ChatGPT. Both connect to the same MCP and secure execution flow. Providers and outcomes remain fictional demo data.

prepare_cadon_context → show_car_loan_bank_optionsget_cadon_demo_linkget_demo_resultActual screenshot of the connected MCP service. Provider cards in the interactive demo come from the actual MCP response.
Website launch access requires a code.Car financing is the flagship MCP walkthrough. The connected plugin also offers fictional lending, savings and payment journeys.
“I need €25,000 to buy a car.”
Connected MCP demo“I need financing for renovations.”
Connected MCP demo“I want to open a savings account.”
Connected MCP demo“I want to send money.”
Connected MCP demoInitiate a controlled insurance workflow.
Platform visionThe connected prototype exposes real MCP tools. This is the actual tool-call request used by the website to prepare a car-finance launch. Initialize the MCP transport before calling tools.
{
"jsonrpc": "2.0",
"id": 2,
"method": "tools/call",
"params": {
"name": "get_cadon_demo_link",
"arguments": {
"scenario": "car_loan",
"market": "BE"
}
}
}The service returns a launch reference and browser URL. Open the URL to begin the fictional flow; use get_demo_result to retrieve its minimal status. No fictional SDK is shown.
Financial actions need an explicit approval flow.
The institution owns its products, requirements and execution.
Return only the information the AI needs to follow the action.
Design principles, not certifications or a guarantee of regulatory compliance.
For financial institutions exploring a new distribution channel, and AI teams that need a controlled path to execution.