Prologue

Life is a succession of events. Your heart beats without you asking, your mind holds barely three or four thoughts at a time, and meanwhile the subscription charge you didn't notice goes through, the medication runs out, yesterday was the deadline, the bank statement doesn't add up. It isn't lack of will. Human attention has limits, and modern life overflows them.

Ten thousand years ago the first farmers faced the same problem. They molded clay tokens, a cone for each measure of grain and a disk for each sheep, because they had more goods than they could track in their heads. That system, documented by archaeologist Denise Schmandt-Besserat, eventually became cuneiform writing. Writing wasn't born to tell stories; it was born to keep accounts.

Then came ledger books, Pacioli's double-entry bookkeeping in 1494, spreadsheets, computers. Every leap in human complexity demanded a new tool to close the same gap: the distance between the complexity of our lives and the finite capacity of our minds.

2,400 years ago Aristotle named the other face of the problem: akrasia, knowing what's right and choosing otherwise anyway. Modern psychology put numbers on it. Your conscious intentions only predict between 18% and 23% of what you end up doing. The other ~80% comes from habits, environmental cues, fatigue and the design of your surroundings. Your intention to save competes against a delivery app engineered for one-tap checkout; your intention to eat better competes against a supermarket that places candy at the register.

The problem was never lack of information. Human agency — the capacity to act with purpose — breaks down when the system it operates in is more complex than the mind can manage.

That started to change in June 2017, when eight researchers published "Attention Is All You Need" and introduced the Transformer architecture that enabled ChatGPT, Claude, Gemini and everything that followed. For the first time a machine understood natural language, with its ambiguity and its implicit context.

But a language model on its own can only respond with text. You ask it how much you spent this month and it can't know: it has no access to your bank. You ask the weather and it answers with what seems right, not with what is right now. And sometimes it invents facts with total confidence — what's known as hallucinations.

What turns a model into an agent is one concrete capability: function calling, the ability to invoke tools. An agent receives your goal, observes which tools are at its disposal, decides which ones to use, executes and adjusts as it goes. It doesn't follow a fixed script; it reasons about the situation and chooses what to do. An email filter that moves messages into folders is automation. A system that understands "log the expenses from the Japan trip in this PDF, avoid duplicates with what's already in my account, assign categories and group them under the Japan Trip group with your companions as counterparts" is an agent.

That is Clatri: an AI agent built so you can take control of your finances, your health and your administration & productivity with the ease of a conversation. The name is born from a conviction. The one who is clar, triples. When you have clarity over your money, your health and your time, you don't just move forward. You multiply.

Clatri operates with more than two hundred tools distributed across domain specialists, on top of entities: isolated profiles that can be your personal life, your business, or a team, shareable with whoever you decide. It's the clay token of the 21st century — except that instead of counting sheep, it understands, connects and anticipates.

The pages that follow explain how.