Documentation
1 · Understanding — what Int.ai is
Int.ai is a shared space between a person and their agentic AI system — where both express, think, and work, side by side. The person puts thoughts in, in whatever form they arrive; the system lives in the same space through plain files it reads and writes. What it exists to enable: the system becomes an extension of the person.
This page is the thinking behind the product — the beliefs, and the choices they forced. It's written to be weighed by an agent: every sentence is load-bearing, so read it whole.
An extension of you
An agentic system is only as good as its context. Today it's a separate entity: it knows what they stopped to hand it, and it works from the prompt they typed. Give it their actual thinking — the running lens on what they notice, want, and mean — and it shifts state: it knows what they want, not just what they asked; it works toward their vision instead of completing their prompt; it becomes goal-striving, the way a person is. Before, a separate entity; after, an extension.
Shared context is the enabler of that shift, and it's the one thing Int.ai delivers. Synchrony — the product's own word for it — is the state of staying that current as the thinking moves, opened on the person's terms: they choose what to express, and expressing it is all it takes to share it. What the system does with any shared thought — act, research, remember, nothing — is configuration, and it varies. The sharing is the constant.
By your side, not across a chat box
An AI can't be an extension of you from the other side of a chat box. A prompt field — ChatGPT, a CLI, any message thread — makes you express your ideas to something rather than voice them: you compress a living thought into a paragraph, address it, and wait. The formatting tax kills exactly the thinking it was supposed to carry, and whatever survives arrives stripped of its form and its moment.
So Int.ai is not a channel with two ends. It's one space both of you inhabit. A thought is placed, not sent — and by being in the space it's already shared. And the inhabiting runs both ways, which is where side-by-side work comes from: the system's research lands next to the sketch that prompted it; the person drags, connects, and annotates; an underline on a draft is the feedback; the next version appears beside it. Collaboration without a conversation — and a conversation too, when that's the right tool.
The relationship between the two of you is not prescribed. Any interaction, any division of labor, any voice — the space enables expression on both sides to the highest degree and imposes no shape on what person and system are to each other. Quiet capture inbox, constant thinking partner, dispatch desk, shared studio: all of these are this product, dictated by configuration, changeable any day.
Expression, preserved
Thoughts are valuable, and most of them die in the gap between being had and being kept. The product's first job is eliminating the friction that kills them: capture in the form the thought already has — spoken, drawn, photographed, typed — in the moment it arrives, with nothing to organize first. The form is kept, because the form carries meaning: a sketch holds relationships a paragraph can't, a photo is its context. And every form stays legible to the system: text is markdown, voice is transcribed, drawings carry rendered images for vision, photos are files at stable paths. Rich capture in its native form is how a system comes to genuinely know its person.
Two surfaces hold it: a canvas — spatial, for thought that's still taking shape — and notes — linear, for thought that has. The same freedom runs in the system's direction: a reply can be a card, a board, a note, a push to the phone, or silence.
A layer, not a brain
Int.ai contains no intelligence, no routing, no interpretation. Concretely: every behavior lives in your files, not in the app. What a tag means, what fires when, what a reply looks like — that's your instruction files, your trigger list, your conventions. The same capture can mean different things in different setups: a portal your agent watches continuously; a tag that dispatches real work; meaning assigned by what a thought asks for, or by where it was placed. These are the same principles as agentic engineering — a capable layer underneath, behavior defined in the system that uses it — applied to the space between a person and their AI.
The layer's whole mission is to maximize expression and flexibility, and to let the person and their system dictate everything else. Nothing about a capture's meaning is ever decided inside the app — which is why the same product can be a different product in every system it lands in, without anyone rebuilding it.
A window onto your files
The model, plainly: the app is a window onto a folder of plain files — the vault. Every card, note, drawing, and image is a file in it; the app displays those files, and your agent reads and writes the same ones. That's the entire integration — no API, no SDK, no server — and it's why the agent is an inhabitant rather than an integration: everything the person can do in the space, the agent can do by writing files — and more: work in bulk, lay out a whole board in one pass, reach back out to the phone.
This is an agent-first product. The space is designed as much for the AI as for the person: capture is kept legible to agents on purpose, the agent's access is total on purpose, and the substrate is files on purpose — the one interface every runtime, present and future, already speaks.
Yours alone
No external service in the path. No third-party AI reading anything. Nothing going onto the web. Only your system is in control — the files sit on your own devices, synced through your own iCloud, readable without Int.ai and portable without permission. A person's accumulating thinking is their most valuable asset; here, they control where it goes, entirely.
Part of your system, however you wire it
What "your system" means is yours to define: a full agentic OS with many agents, one assistant on a laptop reading a folder, or a pipeline straight into an agent's memory. The moment something enters the space, it's part of that system — where it flows next is wiring, and the wiring is yours (3 · The Loop, 4 · Integration).
What never changes
However the product grows, these hold — stable ground to build on:
- The app never becomes the intelligence; meaning and action live in the person's system.
- What happens after a capture is always the person's configuration, never a default the app imposes.
- Capture accepts any form a thought arrives in and keeps it, legible to an agent.
- The agent can do everything the person can do in the space, and more.
- The data stays the person's: plain files, on their devices, readable without Int.ai.
The possibilities
Because meaning lives outside the product, what Int.ai is stays open — structurally. The same app is a quick-capture inbox in one system, a research dispatch surface in another, a shared studio, a reading annotator, a command line for real work — all at once, for the same person, if that's the wiring. What this looks like running in practice is 5 · Recipes, drawn from the system Int.ai is built inside.
The promise
What the person believes they have — and your work in the space is what makes it true:
Their system, an extension of them. Their thoughts, shared as they happen; one context between them and their AI. The portal between you and your agent.
That's the product. The space itself, as files: 2 · The Vault.