“I want to insure my car.”
The rating engine requires
34 to 45 steps to a signed policy
What it costs you
Sales. Every step is a chance to leave.
Verged Mundaka
A language layer between people and machines.
People say what they need. Mundaka works out what the system behind it requires and brings back the system's answer.
Verged Mundaka is a language layer between people and machines. A caller states a task in its own terms. Mundaka works out what the target system's interface requires, asks back where the task is unclear, and returns the system's own answer with its basis. The decision stays with the person.
Why are so few people using the digital services you have built so well?
Because every interface leads people toward the system, and none starts from what they want. The translation in between is left to them.
“Who is showing cybersecurity here?”
The platform filters by
the category IT security
What it costs you
Contacts. The visitor leaves without meeting the right exhibitor.
“We want to digitise our production.”
The programme checks
SME status, de minimis position, funding area
What it costs you
Uptake. The funding misses the companies it was meant for.
In all three cases the answer is in the system. People who cannot reach it give up or never start. Better design changes little. The requirements come from tariffs, categories and funding rules, and those stay.
Often the load sits one level higher. Someone who moves house deregisters, registers again, updates their ID card and applies for a nursery place, each time in a different system. A single service case leaves a call record in telephony, a ticket in case management and a transaction in the system of record. Here too, the person does the translation.
How does Mundaka work?
Take the task as stated.
The input can be a sentence, a website address or an email address, in the words of whoever is asking.
Work out what the target system requires.
Mundaka knows the categories, required fields and derivation rules of the target system and derives what its interface needs.
Ask back where the task is unclear.
If a detail is missing or ambiguous, Mundaka asks instead of guessing.
Return the answer with its basis.
The answer comes from the system itself. Every derived value carries its source and the rule applied, and every step is logged.
How is Mundaka different from a chatbot?
Mundaka has no interface of its own. That follows from the category. People use Mundaka in the assistant, chat or website they already have, with no new app and no new account. The organisation behind it embeds Mundaka through MCP or an API into the surface it already operates. Where no API exists, Mundaka docks onto the web surface. A chatbot can use Mundaka as its backend. Mundaka is not a chatbot.
Connecting to systems is no longer the hard part. The Model Context Protocol (MCP) and the MCP servers vendors now publish have turned it into common infrastructure. And the assistants that system vendors ship cover their own system only.
What Mundaka brings is the knowledge that belongs to one target system: the categories it uses, the fields it requires, the rules by which values are derived, and when to ask back. That knowledge sits in no training data. Mundaka works where a system has no assistant of its own, and where one request touches several systems.
Mundaka changes no target procedure and makes no automated decision.
Where does Mundaka come in?
Mundaka for Events
Answers visitor questions about exhibitors, sessions and products from the event platform's live data, inside the chat the organiser already runs.
In production behind the Smart Country Convention chat since August 2026.
Mundaka for Public Administration
Designed to derive the formal fields a funding programme requires from what applicants know about their project.
Funding programmes come first. Submission and register connection are in development.
Mundaka for Insurance
Designed for taking out a policy and reporting a claim, the two points where customers meet an insurer's systems.
The aim is digital sales and claims journeys that customers finish instead of abandoning.
Coming soon
How is Mundaka introduced and billed?
Mundaka is introduced per backend. In the setup we connect the layer to the target system and to the frontend that sends requests, and we store the rules of the domain. After that we operate the layer.
Billing runs per result where a system event proves the result, otherwise per call. Vendors who embed Mundaka in their own product take an OEM licence. We never bill on an outcome someone else decides, such as a grant awarded. What counts as a result is defined in each contract.
What does Mundaka rest on?
Mundaka rests on Verged's guidelines for ethical AI. The layer makes no automated decision. The decision lies with the person asking or with the body responsible.
Explainability has two parts in Mundaka. Every derived value carries its source and the rule applied. And a change in the input shows its effect on the output. What these principles look like in a given domain is settled per application.
Frequently asked questions
Is Mundaka a chatbot?
No. A chatbot is an interface people ask through. Mundaka has no interface of its own. It sits underneath and translates what comes in through a chatbot, a portal or an AI assistant into what the target system requires. A chatbot can use Mundaka as its backend.
How would we embed Mundaka in our own software?
Through MCP or an API. Mundaka has no interface of its own. The language layer works in the surface you already run, whether that is your portal, your app or your assistant. What Mundaka brings is the knowledge of the target system: which fields it expects, which rules apply, and what has to be asked back before anything is sent.
Does Mundaka make decisions?
No. Mundaka works out what a system requires and returns the system's answer. The person asking or the responsible body decides. Every derived value can be traced to its source and rule, and every step is logged.
Which systems can Mundaka connect to?
Systems whose interface is rigid or complex for people and that return a live answer, meaning a record or a system event rather than a document. The connection runs through an API where one exists, otherwise through the web surface. If a document index can carry the answer, a chatbot with a knowledge base is enough.
How does Mundaka differ from an event platform's own assistant?
Large event platforms ship assistants that answer over their own platform's data. Mundaka works where a platform has no assistant of its own, and on questions that cross an organiser's systems: event platform, website, ticketing and hall plans.
Which system asks too much of your customers?
Tell us about the target system and the questions that fail against it. We will tell you whether Mundaka fits there and what the setup involves.