What Is a Snappy Credit?

In my last Founder Diary, I wrote about Tokenmaxxing and why I think the important question is no longer how much AI we use, but how much value we can get from every token we use.

That also explains a small but important design decision we recently made in SnapLedger:

Snappy Credits.

Behind almost every serious AI task today are tokens.

Different models consume them differently. Reasoning costs tokens. Searching, reading documents, evaluating results, running loops and trying again all consume tokens.

But I don't think a small business owner should have to care about any of this.

If you want Snappy to help find better suppliers, you shouldn't have to think:

Which model should I use? How large should the context window be? How many tokens will the reasoning consume?

You should simply say:

“Help me find better suppliers for this product.”

And let Snappy figure out the rest.

This is why we introduced Snappy Credits as a simple abstraction over the increasingly complicated AI infrastructure underneath SnapLedger.

Every paid SnapLedger subscription comes with 5,000 Snappy Credits every month.

You can see your usage and add more credits when needed from the Snappy Companion Settings.

But what does 5,000 credits actually mean?

This is where our previous discussion about Tokenmaxxing becomes important.

A good sourcing task might ask Snappy to understand what you are looking for, design a sourcing strategy, search across different sources, evaluate potential suppliers, reject poor matches, compare the remaining candidates, and perhaps run another loop to improve the result.

A marketing task might start with:

“I think companies like this could be my customers.”

Snappy can turn that idea into a target profile, develop a search strategy, test it against real prospects, evaluate the quality of the results, refine the strategy, and eventually turn it into something that can actually support an outreach campaign.

These aren't simple chatbot questions.

They are work.

Yet a well-designed sourcing search or marketing task may consume only several to a few dozen Snappy Credits.

So 5,000 monthly credits can go a surprisingly long way for a typical small business.

And if a business becomes a very heavy Snappy user, additional credits can simply be purchased as top-ups. The design keeps the monthly allowance and additional balance in the same shared Snappy wallet, while top-up credits do not expire.

But there is a more important idea behind this.

We don't want to maximize how many credits Snappy consumes.

We want to maximize what Snappy accomplishes with them.

This is where much of the invisible engineering inside SnapLedger matters.

We can route simple tasks to efficient models and reserve more powerful models for genuinely difficult problems — an approach that is already part of our Snappy design.

We can use Loop Engineering so that an agent doesn't endlessly think, but works toward a defined acceptance goal.

We can retrieve only the relevant business information instead of repeatedly feeding an entire company history into a model.

And, perhaps most importantly, Snappy keeps learning about your business.

It knows your transactions.

It reads your invoices and contracts.

It learns about your customers and suppliers.

It understands the documents you give it.

It remembers previous conversations and decisions.

This means that as SnapLedger builds better context around your business, Snappy should spend less effort repeatedly asking:

“Who are you?” “What does your company do?” “What are you trying to achieve?”

And spend more of its intelligence actually working on the problem.

That is our version of Tokenmaxxing.

Not more tokens.

Not more AI for the sake of AI.

But better context, better models, better loops and better memory — so that the same amount of AI can accomplish more useful work.

I think this is another important responsibility of AI-native software.

The technology underneath AI is becoming more sophisticated incredibly quickly.

The experience for the user should become simpler at exactly the same time.

A small business owner shouldn't need to understand tokens, context windows, model routing or Loop Engineering.

They should understand one thing:

I have Snappy Credits. What can Snappy get done for me?

Our job is to make the answer:

More and more.