Edge Parser

Invoices & reconciliation — for owners and practices

Reconciliation that shows its work.

Throughout the month, receipts and invoices file themselves into the archive. At month end, your AI assistant works the bank statement line by line — and Edge Parser hands it what the job needs: the receipt for every line, the history of how each supplier was treated, the document to attach. Every entry lands marked for review. You approve. That's the whole job.

From CloudMailin · Works with Xero, QuickBooks & FreeAgent · Every entry marked for review · Flagged, not fudged

See a month

A bank statement, worked line by line.

This is your AI assistant at month end, working from an archive where every receipt is already filed. Every match explains itself. Every doubt gets raised, not buried.

March — bank reconciliation

23 of 25 matched
  • GOOGLE ADS — £412.16

    Matched: Google Ads invoice · Mar 2026

    Filed · for review

    Why: Same category and VAT as the last 14 months.

  • HEROKU — £61.33

    Matched: Heroku receipt · Feb usage

    Filed · for review

    Why: No PDF exists — the email itself is the receipt, attached as an image.

  • OPENAI — £142.10

    Matched: OpenAI invoice · Mar 2026

    Filed · for review

    Why: Overseas software supplier — reverse charge, the way your accountant expects.

  • AMZN MKTP UK — £38.94

    Matched: Amazon receipt · USB-C cables × 4

    Filed · for review

    Why: Invoice dated 28 Feb; the card settled 3 Mar. Matched to the document, not the date.

  • ACME SIGNAGE — £240.00

    First time this supplier has appeared

    Waiting for you

    Why: No history to copy, so it asks once — and remembers.

  • TFL TRAVEL — £8.10

    No receipt found

    Flagged

    Why: Flagged, not fudged. You decide.

…and 19 more like the first four. 23 filed as drafts with receipts attached · 2 waiting for you · 0 guessed.

Illustrative month. The method is real — it's how our own AI assistant does CloudMailin's books, working from the archive.

How the deciding works

Your AI does your books like someone who's done your books before.

Edge Parser doesn't do the reconciliation — your AI assistant does. Edge Parser's job is making each of these habits possible: the receipts, the history, the attachments, all one question away.

History gets checked first

Before categorising anything, your AI looks at how that supplier was treated last time — same category, same VAT, same wording, month after month. The archive is where that history lives, findable in one ask.

The invoice gets matched, not just the amount

A card charge lands days after the invoice it pays. An annual renewal isn't twelve monthly ones. Your AI matches the actual document — because the archive can hand it the actual document, not a guess about a bank descriptor.

VAT follows who the supplier is

A UK supplier books at standard rate. An overseas software supplier is reverse charge. Your AI decides by who the supplier actually is — and the invoice in the archive is what says so.

Doubt gets flagged, not fudged

New supplier, odd amount, history that doesn't agree with itself — your AI stops and flags rather than guessing. The archive makes "no receipt exists" a checkable fact, not a hunch.

The part that matters

Nothing posts without you.

Your AI does the tedious 90% — finding, matching, attaching, drafting. The judgment stays with a person. That's the setup we run on our own books, and the only one we'd recommend.

  • Drafts, never postings. Every entry your AI creates arrives marked for review — approving is deliberately a human job.
  • The receipt comes attached. No bank line books without the document that justifies it.
  • You work a queue, not a keyboard. Read, approve, amend — the final say is yours, entry by entry.

The other half

The receipts are already in.

This works because the archive is full before month end arrives. Suppliers email one address; paper gets snapped and sent in; everything files itself the moment it lands. Even the receipts that never come as PDFs — the ones that are just an email — are found and attached.

What goes in, and how →

The books you already keep

Xero · QuickBooks · FreeAgent

Whatever you run, there's nothing to switch and no plugin to install. Your AI assistant reads the archive on one side and works in your accounting software on the other — the same way you would, with your permission, every entry as a draft for you to approve.

Our own books live in FreeAgent, and it reconciles them every month. Xero and QuickBooks work the same way — because the AI works your software rather than plugging into it.

Not a demo — our own books

The first customer was our own bank statement.

Edge Parser is a CloudMailin product, and before it had a landing page it had a job: CloudMailin's own books. Every receipt and invoice the business gets files itself into the archive — the Google Ads invoices, the hosting, the software subscriptions, the receipts that only ever existed as an email.

At month end, our AI assistant works the statement in FreeAgent — pulling each receipt from the archive, checking each supplier's history, drafting the explanation with the document attached, marking every entry for review. A person reads down the queue and approves.

Month end used to cost us a weekend. These days it costs a coffee.

For bookkeepers & accountants

Your review queue arrives pre-worked.

You don't bill for typing in receipts. You bill for judgment — the category call, the VAT edge case, the thing that doesn't look right. With every client document filed in Edge Parser, your AI assistant can hand you a queue where the first 90% is already done, documented, and defensible: every entry consistent with the client's history, every source document attached, every doubt flagged instead of fudged.

Your name stays on reviewed work

Everything arrives as a draft, marked for review. Nothing carries your sign-off until you've actually signed it off.

Ready for the enquiry letter

Every entry ties back to its source document — already attached, already consistent with how the supplier was treated before.

It escalates like a good junior

New supplier, inconsistent history, amounts that don't tie out — flagged to you with the context, not guessed at and buried.

Today, Edge Parser runs one set of books per account — most practices start by running their own books on it. Multi-client support is on the way, and the early practices are shaping it.

Talk to us about your practice

Straight answers

The questions your accountant will ask.

Does Edge Parser do the reconciliation itself?

No. Your AI assistant does the reconciliation — Claude, ChatGPT, whichever you use. Edge Parser is the half it can't do alone: every receipt archived, findable, and attachable the moment it's needed.

Which AI does it work with?

Claude, ChatGPT, or any assistant that speaks MCP — the open standard the major assistants already use. The wiring details are on the page for techies.

Does anything post automatically?

Not unless you tell your AI to. The method we run on our own books — and the only one we recommend — creates every entry as a draft, marked for review. A person approves before anything posts.

What happens with a supplier it's never seen?

Your AI stops and asks. With no history in the archive to copy, guessing would just create work — you set the treatment once, and it's remembered.

Do I have to change accounting software?

No. It works in whatever you already run. Nothing to migrate, nothing to switch.

Is this a Xero or QuickBooks app?

No — and that's deliberate. Your AI assistant works your accounting software the way you would, with your permission. There's no plugin sitting between you and your books.

Where do the documents live?

In your archive — hosted by us or kept in your own cloud storage, UK or EU, GDPR compliant. Never sold, shared, or used to train anyone's AI model.

What if the receipt genuinely doesn't exist?

It says so. A missing receipt gets flagged as missing — that's the point. You decide whether to book it without one, and the books record that decision honestly.

Where your paperwork lives

Private by default. Yours to keep.

  • Sixteen years behind it. Edge Parser is built by CloudMailin, which has handled email for businesses since 2010 — the same team, the same care, still here.
  • Never sold, never shared. Your paperwork is never sold, shared, or used to train anyone's AI model.
  • Yours to keep. Export everything, anytime, in the same PDF or email format it arrived in.
  • We host it — or you do. Fully managed by us, or kept in your own cloud storage. Either way, industry best-practice security, GDPR compliant, in the region you choose.

Private beta

Stop typing. Start approving.

Tell us a bit about your business — or your practice — and we'll get you set up.

No marketing list. We'll reply personally.