The fastest way to log an expense: just say it.
The reason most people give up on expense trackers is the friction of logging a transaction. Open the app, find the right form, pick a category, type the amount, add a note, save. By the time you've done that twice, you've already lost the receipt for the third one. Voice entry kills the friction: open the app or tap a widget, say one sentence in your normal speaking voice, and the transaction is logged in the right currency, the right category, with the right vendor.
What you can say
You don't need a structured command. The parser handles natural language:
- "Eight hundred rupees on chai at Espresso Lounge."
- "Spent 3,200 on fuel at Shell DHA."
- "Bilkul, paanch sau ka biryani at Student Biryani." (Roman Urdu)
- "USD 49 monthly subscription to Notion, yesterday."
- "1,200 dirhams hotel in Dubai, four nights."
- "Cash, 600 rupees, parking."
What it extracts
- Amount and currency. Numbers and currency words detected together.
- Vendor. Fuzzy-matched to your existing vendors; created if new.
- Category. Auto-suggested based on vendor and your past behaviour.
- Date. "Yesterday", "last Tuesday", "June 4" — all parsed.
- Notes. Anything you say after the core fact becomes the note.
Use cases
In the car
Faisal pays at the pump, taps voice on his phone, says "3,200 fuel at Shell DHA". The transaction is logged before he's back on the road. Compare to the spreadsheet workflow: open laptop at home that night, try to remember the price, fail.
At a meeting
Sara picks up the lunch bill for a client, taps voice, mutters "6,400 client lunch at Xander's, billable to Acme" into her phone. Categorised, vendor matched, billable flag set. She closes the app and the meeting continues.
While walking
Hassan pays the parking attendant and keeps walking. He says "200 parking Dolmen Clifton" into his AirPods. Done. He's already at his next meeting.
How it's different from voice assistants
Siri and Google Assistant can transcribe — they can't parse. The difference between "eight hundred" and "8 100" is everything when it's a transaction. Our parser is built specifically for expenses, knows the numeric quirks of Urdu and English, and disambiguates Pakistani-rupee "PKR 800" from US-dollar "USD 800" correctly even when you don't say the currency.
Related features
- OCR receipts — for the printed slips.
- Multi-currency — for international voice entries.
- AI insights — for what the spoken expenses add up to.
Frequently asked
What languages does voice entry support?
English and Urdu, with Roman Urdu (Urdu written in English letters) auto-detected. Hindi, Arabic and Pashto are in beta.
Do I need an internet connection?
Yes. The natural-language parser runs server-side. Offline capture is queued and processed when you reconnect.
Can I use it while driving?
Yes — voice entry is the safest way to log a fuel expense at the pump. The web app supports hands-free trigger via the system share button on iOS and Android.
How does it know currency?
You say "rupees", "dollars", "pounds", "dirhams" — it captures the currency you spoke. If you omit currency, it uses your home currency.
Does it learn my vendors?
Yes. After the third time you say "Espresso Lounge", it skips disambiguation entirely.
Is my voice recorded?
Audio is processed in-memory and discarded after transcription. We never store voice clips.
Speak your first expense
Free plan allows 3 voice entries per day — enough to test the parser on your real speech patterns.
Start free — no card