LedgerTB - Double-entry Bookkeeping App Built for Humans & Agents

Today I want to show off a couple of apps that I've built.

One of them is called LedgerTV, which is just a trial balance app.

This is a local desktop application, so you can just launch it like you do any of your other desktop apps.

it's meant to be so you can import your transactions, you can import your chart of accounts, you can import, beginning year balances, and then you can kind of run a typical like month end close or financial statement prep inside the app.

So the application itself is pretty basic, honestly.

It doesn't have a bunch of AI features built into it.

It's really meant to kind of have just the bare bones of what you need to do a month in close or a bookkeeping engagement.

So this is the trial balance worksheet.

and this is the dashboard where you can see total access to all liabilities, net income, you can see recent activity.

So what you've done inside the app gets recorded.

And that'll be important later on when I talk about some of the other features of the app.

Scroll down.

You can see your account balances summary.

this is all fictitious data.

This is a sample client that I've built, using Claude and had it import everything in here just for demo purposes.

trial balance worksheet.

You can see the accounts, the beginning balances, the activity.

So if we change this to be from January 1st, let's just do to today.

You can see beginning balances here in these columns.

You can see the activity in these columns unadjusted.

And you can make your ajes, and you can get to your adjusted trial balance.

And then you can compare to final year in this column.

You can easily add journal entries.

You can choose what type of entry it is.

You can adjust your chart of accounts.

You can import your own, chart of accounts.

You can import your, transactions.

You can use a statement CSV.

You can review transactions.

You can see your import history in here and then you can also set, learn patterns as well.

Then once you're done and you get your trial balance like you want it to be, you can create the reports, your trial balance report, your income statement, your balance sheet.

And it's even got a general ledger.

So you can see all the transactions across all of your different accounts transactions.

You can go and kind of review, your transactions.

Here you can do a bank reconciliation.

There's a close process built into it.

You can do a book review, which is where you can just do like an integrity structure sweep on the books to see if anything, is you Know not as it was in prior year or looks off.

and then you see this assistant review here.

So as I said, AI isn't built in to the app, right?

But you do have the ability to use your AI to connect to the app and use this app.

So what I've done there is I've added mcp, functionality into the app.

So you can, you know, use your AI of choice.

read and propose, which means it'll propose say a journal entry, and that'll be labeled for your review so that you can go and review the entry and you can decide whether to review, approve it or change it or deny it, or you can give it full access where it can just go and propose entries on its own.

you can also set your own branding in here.

you can also set client branding and then you can connect an API key here.

So I've done this so that you can use AI to categorize transactions, and set up those, learn rules, by using your own API key and not being reliant on the built in AI into the software because there is none.

Okay, so now I want to show that MCP action using CLAUDE code.

So I have CLAUDE code launch here on the right side via the cli.

And so I want to kind of show right now you can see I only have two clients inside this app, my own books which I've been testing using this app.

And then I have this demo client, Copper, line plumbing as well, just for demo purposes.

But let's say I wanted to add a new client and so I have another client that I've built out that's completely synthetic and made up using Claude that I'll get CLAUDE to set up inside the app.

So if we just go here and say let's set Northline Digital inside of Ledger tv.

Okay, so let's see what Claude did.

Looks like it set up Northline Digital up the chart of accounts set up the opening balances, imported the checking transaction, and imported in the credit card transactions.

So now it's saying, you know, it's drafted journal entries for me to review.

And so let's go in here and let's see.

Okay, we see Northline Digital here.

So let's pull up North Line Digital and see what Claude did.

And as you can see, you can see some activity here on the left side.

Looks like there's a journal entry draft.

There's some import transactions.

Let's go to the dashboard.

Right now you're seeing all zeros, but you do see the activity that Claude did.

And you can see it's labeled AI next to my name.

So that clearly, identifies that AI did this.

But because, the access that I have with Claude is to propose, which means nothing's finalized until I review it and I approve it.

So let's go to the journal entry and review the entry that Claude made.

Looks like it set up the opening balances 12, 31, 20, 25.

We'll just hit approve.

Those look right.

And then now let's go back to the trial balance worksheet.

And we can see those beginning balances now populated.

And you can see the activity where I entered in the beginning balances.

You can see it doesn't say AI because I'm the one that entered those.

Now let's go and see the transactions.

So if we go and look at the import history, we'll see these are the transactions that came in.

Let's go to review.

We'll see that they're staged from our review.

So let's load the ones that were staged.

So we can see there's 72 total transactions here.

None of them have been categorized yet.

Right, because, again, they've just been imported.

And you can use AI to categorize your transactions if you choose to bring in your own API key, which I have done.

So we'll categorize these 72 And you'll see AI is now categorizing those transactions.

So.

So we'll give this a minute.

Okay.

Looks like it's done categorizing.

All transactions have been categorized.

There's five part from our review.

Here's one right here.

It looks like it Claude flagged it as a potential owner's draw.

So we'll just go ahead and change that to be owner's draw.

Keep going down.

Most of these are software Looks right.

let's see if we see another one that's been flagged for my review.

Looks like this one has.

Looks like it's another owner's draw, which we'll go ahead and categorize as such.

Keep going down, see if we see any more.

Looks like there's two here, which look to be two sides of a credit card payment.

So we'll mark them both as transfers and book them to the credit card payable to offset each other.

And then looks like there's one more IRS payment.

This is an S Corp, so we'll go ahead and book that as an owner's draw as well, see if there's any more.

Nope, all those look good.

So we'll hit post and now we'll go back to the trial balance worksheet and we'll see if we change the date here to be April or January 1st.

We'll actually see those transactions being categorized in the activity column next to the beginning balance column.

And now we have our unadjusted, trial balance columns here.

So again, it's a pretty basic application on the surface, right?

But can.

It's really meant to be used with agents and for a professional human to be in the loop while the agent is working inside the app.

So you can see here, become this assistant review.

You can see, the activity of the agent, clearly labeled AI here.

And so it's created all the accounts to get started, it imported all the rows, and you can just clearly see the activity of the agent.

And you can also come to this audit trail and see exactly what, or who did what, whether the agent did it, whether I did it, and just kind of easily see this in one ui, which I makes it for a cool way to work alongside agents while still staying in the loop.

Another cool feature is that you can save.

You can create a backup of your books and keep it locally, and then you can go and restore your books if something were to happen, and you can pull it straight from your local.

So again, everything lives locally unless you hook up a mcp, or you bring your own API key, in which case, if you use either one of those, then whatever the agent does, it'll, you know, use that information to send it to its server, whatever software you're using or whatever model you're using.

but other than that, you can toggle both of those off and have everything live on your local machine.

Okay, let's go back to Claude and Say, okay, all transactions have been approved as well, as the journal entry.

And we'll see what Claude comes back with.

It looks like Claude identified some other potential journal entries that we should make.

So let's go and see what those are.

We go to journal entries.

See, there's two draft ones here.

Let's go and review what they are.

Looks like one of them is retainer revenue, reclassify into unearned revenue.

So for work that hasn't been completed yet, so we'll just approve that one.

The other entry is what looks to be personal charges.

So, you know, in this situation, these could very well be business expenses.

we'd have to ask the client whether, they're personal or their business.

For right now, we'll approve these being personal expenses.

So we're going to approve, and we'll go back to the dashboard again.

You can see that activity approved by me.

You can see, the updated account balances.

And then now we'll kind of close this and export out the financials.

Okay, so it looks like Claude went ahead and created the CLOSE package, You can see it here.

It looks like there's a summary page.

There's the income statement, revenue and expenses balance sheet, final trial balance, all the transactions that came in, and also the adjusting journal entries that were made.

And so again, Claude, also ran an integrity sweep over the books.

You can see it here.

Seven checks, zero findings, every target figure hit, everything's in balance, and went ahead and created the CLOSE package.

And this exported out to my desktop.

Because if you go into actually data safety, you can set where you want, things to be exported.

So you can decide which folder you want the export package to go.

In this case, I just have it on my going to my desktop in this Demo Documents folder here.

But, you know, again, a very basic application, right?

And this is just a V0 built entirely with Claude.

and I'm also running it with Claude.

But, what I didn't want to do was just bolt on a bunch of AI features into the app.

Instead, I wanted to build something to where I could work alongside an agent and also see what the agent is doing inside of the app and then still have the level of control that I want to have over what the agent does and easily review what the agent does and either approve it, either deny it, and kind of, again, work alongside the agent while still staying in the loop.

So this is again, a desktop application.

You can Just go and download it yourself.

So you just go to ledgertv.com you can hit download.

It's also open source, so you can go to GitHub and you can inspect the code yourself, you can fork it, you can branch off of it, you can basically build it out to however you want to build it yourself.

It's completely open source.

That's also kind of my goal with this and really where I see things headed accountants and other professionals building their own software, but also working on that software with other professionals and other accountants and really building out applications to really suit their needs, rather than relying on a pre built software that may have features that you don't need or not have features that you do need.

So again, download it for Windows, you can download it for Mac, you can go to GitHub, and you can go and review the code and so you can hand this to your IT person or you can go hand it to your Claude and say, claude, go review this repo.

What is it?

Is it safe to download?

What does it do?

How do I use it?

And everything's laid out here in the readme.

And so Claude should be able to easily identify what the app is, what the purpose of the app is, how to get it set up, and kind of what to do with it.

So please go check it out, let me know what's broken, let me know what other features you'd like for it to have and let's work collectively to build our own software.