Most finance apps treat AI as a chat box bolted onto the side. Spendly does the opposite: it exposes your finances as a proper tool surface, so the assistant you already use can work with them directly.
Spendly speaks MCP
Spendly ships an MCP integration — the Model Context Protocol, the standard for giving AI assistants real, scoped access to a system.
Connect it to Claude and you can simply ask:
“Here’s my bank statement PDF — add these to Spendly, categorised.”
“What did I actually spend on eating out over the last six months?”
“Set up budgets for next month based on what I averaged since March.”
The assistant reads your accounts and categories, then writes transactions in bulk — up to a hundred at a time — as if you’d entered them yourself. No copy-paste, no CSV wrangling, no fighting an import wizard’s column mapper.
What it can do
| Read | Accounts, categories, transactions, budgets, budget history, investments, dashboard |
| Write | Create and update transactions, bulk-create up to 100 at once, create accounts, set budgets in bulk |
Access is scoped, not total
The integration authenticates through OAuth with separate read and write scopes. An assistant granted read access can summarise and analyse but cannot change anything. You grant write access deliberately, and you can revoke it at any time without touching your password.
This is the part most “AI finance” features get wrong. Handing an assistant your login is not integration, it’s a liability. Scoped tokens over a defined protocol are how this should work.
The budget coach
Separately from anything you connect, Spendly has a coach built into the product.
It reads your recorded history — this period’s pace, the forecast, and the trend across recent locked periods — and when a budget is meaningfully out of line, it proposes one concrete change with the reasoning attached:
You’ve averaged €430/month on groceries over the last six periods against a €350 budget. Raise it to €430?
Accept it and the budget updates. Dismiss it and it goes away.
Two deliberate design choices:
It only speaks when it has something to say. Suggestions clear a threshold before they appear — a budget has to be meaningfully over pace, or well under it across several periods. No daily nudges, no notification theatre.
Every suggestion is traceable. It tells you which periods and which figures produced it, all from transactions in your own account. You can check the arithmetic. Nothing is inferred from an outside dataset or a model’s guess about people like you.
What-if, before you commit
The what-if sandbox lets you drag budget amounts around and immediately see the downstream effect on your cash flow and year-ahead projection — then apply the result to your real budgets if you like it.
It’s the “can I afford this” question answered before you find out the hard way.
AI budget drafting
Rather than building budgets category by category, have them drafted for you.
Spendly reads the transactions you’ve recorded and proposes a full set of budgets — each with a plain-language rationale so you can see where the number came from:
You averaged about €430/mo on groceries over the last few months.
Steady €260–300/mo dining out; €280 leaves a little headroom.
Mostly rideshare — €120 covers a typical month.
Accept the ones that look right, adjust the ones that don’t, dismiss the rest. It turns the worst part of starting a budget — guessing at numbers you don’t have — into a review step.
Why this combination is unusual
Plenty of tools have a chatbot. Very few give an assistant genuine, scoped, structured access to your financial data — and fewer still pair it with the planning depth to make the answers worth having.
Spendly does both: envelopes, sinking funds, debt payoff, cash flow and safe-to-spend, across 49 currencies — all of it readable by an agent you control.
Open the app or see pricing.