How to Connect Rivhit (רווחית) to AI and Automations via Its API

If your business issues invoices and receipts through Rivhit, you are sitting on structured financial data that AI can read, summarize, and act on. You don't need to migrate off Rivhit or copy data into a new system. You connect to its API, pull the documents and customers you already have, and route them through automation. The good news: Rivhit's API is real and reasonably documented. The honest part: it behaves like an older Israeli enterprise service, so the integration work is in handling its quirks reliably.

How the Rivhit API actually works

Rivhit exposes a web API (the RivhitOnlineAPI service) with methods for creating and reading documents, receipts, and customers. Authentication is token-based: you take the API token from your Rivhit account, call the connect endpoint to open a session, and then call methods like Document.NewExtended to create an invoice or Document.Details to read one back. Requests and responses are JSON. There is no OAuth dance and, importantly, no native webhook system — Rivhit will not push an event to you when a new invoice is created.

That last point shapes every integration. Because there are no webhooks, you build a small scheduled job that polls Rivhit on an interval — every few minutes, or hourly — asks 'what changed since last time?', and processes the new documents. It is not elegant, but done properly it is completely reliable.

1. ApiToken.Connect(token)        -> session
2. Document.List(since=last_run)  -> new invoices
3. for each invoice:
     -> send to AI for extraction / classification
     -> write the result somewhere useful
4. save last_run timestamp

What you can actually build

• Auto-filing: every new invoice is read by AI, categorized (rent, suppliers, marketing), and pushed into a Google Sheet or your accountant's folder with a one-line summary.
• WhatsApp answers from real data: a customer asks 'did my payment go through?' and an AI agent checks the matching receipt in Rivhit and replies — instead of someone digging manually.
• Cash-flow alerts: a daily AI digest of unpaid invoices over 30 days, sent to you on WhatsApp or email, with the chase message already drafted.
• Reconciliation: match Rivhit receipts against bank or payment-processor lines and flag the mismatches an AI couldn't auto-resolve.
• Smart issuing: dictate or paste a job description, and AI drafts the correct Rivhit invoice (line items, VAT, customer) for you to approve before it's created.

The AI model is the cheap, easy part now. The real work is the boring layer: tokens, polling, VAT edge cases, and never corrupting a financial document.

No-code or custom code?

Be honest about scale. If you issue a handful of invoices a day and just want them copied into a sheet with a summary, a no-code tool like Make can call the Rivhit API directly and you may never need a developer. But the moment you need real volume, private financial data that shouldn't pass through third-party servers, polling that doesn't miss or double-process documents, or logic that touches VAT and accounting categories, you want custom code that owns the reliability — because a wrong number on a tax document is not a bug you want to discover at year-end.

I build exactly this kind of integration: a small, reliable layer between Rivhit and AI that runs quietly and doesn't touch your books in unsafe ways. If you're looking to hire a developer to connect Rivhit to AI or to automate your invoicing end to end, tell me what you're issuing today and what you wish happened automatically — and I'll tell you honestly whether it's a no-code afternoon or a proper build. Either way, you'll get a straight answer.