The agents take the client from home search to application. The verdict is either “buy now” or a plan to get there. Either way you get an applicant. The loan and the client stay yours.
Built for Europe from day one. The property models were born in Spain, the finance side was proven in the Czech Republic.
But no one reads both sides together. Marb is the purchase underwriter: one verdict on whether the flat is worth the price and what the listing left out. You decide the loan.
One purchase underwriter, four agents. Each covers one moment of the decision.
Most people don’t know what they’re actually looking for. The Scout brings their life into it: work, school, budget. It combs the whole market and returns only the flats that fit right away.
→ points of interest + area prices → market matchThe Analyst reads accounts from every bank and self-employed statements, runs DSTI and a stress test, and builds a budget so the client doesn’t end up living on plain rice. The bank decides the loan.
→ transactions from every bank · DSTI · stress testOne year, or five? The Navigator breaks it down into months: what to save and when to apply for the mortgage. Then it counts down to the client’s moving day.
→ accounts across banks · your move-in dateThe listing shows pretty photos. The Inspector shows what it left out: condition and repairs from the photos, owners’ association debts, planned construction, floods and noise from official sources. The client knows what they’re getting into before they fall in love.
→ photos · owners’ association · construction · floods · noiseThe agents run through the market and the client’s data and return one verdict over the API: buy now, or a precise plan to get there.
Only now can this be computed: PSD2 opens the account, AI reads the home from photos and the land registry, a model holds each bank’s rules. Our own data, not scraped portals.
They see the client across all their banks, not just one account with you.
With every completed case they get sharper. Every verdict is explainable, auditable and ready for your model-risk team.
With the client’s consent, the agents read money flows across all their banks via PSD2. And turn them into steps that end at your product.
From transactions it assembles real income: main, side, loan payments and BNPL. Everything that decides a mortgage, it knows before the client applies.
client K-2481 · income 62,400 · side 8,000 · probation to 3/27 · BNPL 2,490/mo
You see the client the way your scoring will. Just months earlier.
Food delivery at 4,200 a month is 50 thousand a year toward the down payment. The agents show the client where the money for a home can realistically come from. With numbers, not lectures.
insight · food delivery 4,200/mo → +50,400/yr toward down payment
A client who reaches the down payment sooner is your applicant sooner.
The probation ends, the BNPL gets paid off, the down payment grows. The client sees their progress every month. You see the day they will be ready to apply.
recap 6/26 · readiness 58 → 64% · application possible from 2/27
The pipeline fills itself. Every client comes with a date to call.
The client won’t buy? The agents see from the photos and finances why, and what makes sense instead: renovating their current home, cutting energy costs, a new provider. You get a trigger for the product that fits.
trigger · own flat: facade + windows → green loan · est. 480,000
A renovation loan, a green loan, a new energy provider. Every verdict can become business.
From the photos, Marb identifies every room and its real condition: whether “renovated” means a new bathroom or just repainted walls. And it calculates what the repairs will swallow.
The agent goes through the photos of every listing. It sends an inspector to the one the client picked.
For the chosen property, with the client’s consent, it books an independent inspection. The condition is verified before you lend against it.
You keep the front-end, the data and the relationship. We never go to your client directly. It fits into your stack, it does not rebuild it.
Call the endpoint, get the purchase verdict. Drop it into your AI assistant or app, in your design and under your brand.
Don’t want to build a frontend? Deploy our app in your colours and both the broker and their client work in it.
Connect Marb as an MCP server and work with the client in the AI you are used to. Claude, ChatGPT or your own agent.
Control what matters to your business with built-in guardrails, secure access, and seamless data management. The credit decision never leaves the bank.
People ask ChatGPT about money. 200 million of them every month, without your data and without the truth about the property. Marb puts you at that first question before the competition wins the client.
OpenAI, 5/2026
People let AI explain the purchase. The money goes to the bank standing at the first question. The agents deliver the verdict, the loan and the client stay yours.
Be there for the first question
