Works without internet
On your premises
On Kokonut Station, or with your own local inference engine.
- FYSK
- Ollama
- LM Studio
Accounting standardisation for AI.
Kokonut Wave makes sure every AI works from the real context of your client engagement.
The same document, three AIs, one entry.
Today, each AI has its own interpretation of the client file and of your question, and may return a different entry. Kokonut Wave keeps them in line by drawing on your engagement's drop, for a consistent result that stays stable over time.
- Documents as PDF or photo
- Entries always balanced
- One VAT line per rate
- Entries to check flagged
Compatible with every AI on the market.
- ChatGPT
- Claude
- Gemini
- Mistral AI
- Qwen
- DeepSeek
- Infomaniak
Cloud
Drop in your .kdf file, choose your AI.
Any AI on the market, your choice
What Wave accepts as input.
Two families of files, and the usual formats for each.
Supporting documents
Every document comes back as an entry.
- Photo or scan (JPG, PNG, HEIC…)
Bank statements
Every transaction comes back as an entry.
- CAMT (ISO 20022)
- Bank CSV
- JSON sent by your software
Frequently asked questions
Why don't two AIs return the same entry?
Each AI reads the document and the client file its own way: a different expense account, VAT not split out, an old rate. Without a framework, the result changes from one AI to the next, and sometimes from one day to the next.
What is the .kdf file for?
It is the engagement's accounting profile, created by Kokonut Drop. Wave uses it for every document, with technical tools to interact precisely with any AI.
Can a software publisher integrate Kokonut Wave?
Yes: any accounting software can integrate the services offered by Kokonut Intelligence, with our upcoming SDKs (JS, Laravel, Python, .NET, cURL).
It all starts with a drop.
Kokonut Drop packs each client file's profile into a single file.
.kdf file
Kokonut Drop File
- RAG
- QLoRA
- Vector database
Ready for Kokonut Wave