Manifesto
Moneta is a personal finance assistant that answers by voice or chat. You talk to it like you'd talk to a friend, and it logs, categorizes and shows you the chart. No spreadsheet, no form to fill, no clicks to hunt.
The idea is simple: a good finance tool is one that gets out of your way. You open it to ask, it answers, you close it. Your routine goes back to normal, only now the expense is logged, the category is right, and the chart is ready whenever you want to look.
What Moneta does
Two pillars that stand up without a complex UI. You talk, it acts. You ask, it shows.
A chat that logs
Real-time voice
OpenAI Realtime plus ElevenLabs. Minimal latency, natural voice, no synthetic robot tone. Speaks and listens at the same time, without waiting for the sentence to end.
Actions mid-conversation
The assistant executes: logs a transaction, adds an account, generates a chart, changes a category, all while talking. It doesn't lose the thread.
Chat too
Not in the mood to talk? Type. Same intelligence, same actions, same context. Switch whenever you want.
Your personality
Pick the tone (formal or laid back), voice and avatar. The assistant learns how you like to be addressed, then sticks with it.
A number that explains itself
Charts on demand
"How much did I spend on Uber in June?" turns into a chart on the spot. Bar, line, pie, heatmap. The assistant picks whatever fits best.
KPIs always at hand
Total balance, monthly income, monthly expense, net flow. Cards you glance at for two seconds and know where you stand.
Auto categorization
Log the transaction and the assistant already suggests a category based on context. You approve or adjust, it learns for next time.
Multiple accounts, one view
Checking, savings, credit card with cycle (closing and due date). Consolidated view without adding it up in your head.
How it works
Four steps that run in seconds, without breaking your flow. No intermediate screens, no three-step confirmations.
"Spent 40 on Uber", "Got paid", "How much is left this month?". Natural language. No form, no code.
The LLM parses intent, extracts value, suggests category, validates context. If data is missing, it asks. It doesn't make things up.
Transaction saved in the right account, right category, right date. Syncs between web and mobile. You only confirm if something feels off.
"How much did I spend on Uber in June?" generates a chart instantly. Bar, line, pie. The assistant picks. You just look.
What's next
A roadmap slice with spec, ticket and code already in motion. Not a wishlist. It's the next release.
Fixed salary, variable salary (with monthly override) and fixed expenses as templates. The assistant materializes the month's transaction with one sentence, no typing the amount every time.
"Can I afford this?" turns into an answer based on expected income, fixed expenses and a rolling three-month average. Simple arithmetic, natural explanation.
Native mobile notifications before the card closes, when a category blows past the usual, or when the assistant spots something worth flagging.
Generic OFX plus parsers for the popular banks plus LLM fallback for the exotic ones. Drop the file, it reads and categorizes. No typing.
Capabilities
Voice, chat, charts, personality and security orbiting the same assistant. You don't pick a channel, you use whichever fits.