Context shouldn't live in ten tabs.

Email, chat, issues, calendar, and the open browser tab each hold a fragment of what you need to know. The cost isn't just switching apps—it's re-building context every time you answer a question, write a follow-up, or prep for a meeting. Most “assistants” either ignore that fragmentation or ask you to paste your life into yet another notebook.

We started Noet because the interface that makes sense for this problem is the desktop: close to your shortcuts, able to hold permissions thoughtfully, and honest about when it's reading your tools versus guessing. We believe an assistant should feel like it's on your side of the screen—not a billboard trained on engagement.

That conviction shapes everything: we bias toward retrieval over invention when facts live in your integrations, toward explicit consent for anything that could surprise you, and toward sustainable economics so we're not forced to monetize attention or sell your data to keep the lights on.

In short, we're building Noet for people who do real work across real systems—and who are tired of pretending a single chat window can replace the graph of obligations behind it.

Noet is a macOS desktop application built to sit alongside the tools you already use. You connect integrations with standard OAuth flows; credentials stay where they belong (including platform keychains where applicable)—not scattered across ad-hoc scripts or browser extensions alone.

Under the hood we combine local persistence for responsiveness and privacy-sensitive work with sync where it helps you move between machines. Search and memory are designed so answers can cite the threads, tickets, and events they came from—so “Ask” feels like surfacing what you already have, not improvising a story.

For browser context, the companion extension is deliberately narrow: capture what you choose to save, with clear scope—not silent scraping of your entire history. On the AI side, we route operations to appropriate models and enforce budgets so costs stay predictable; when limits bite, we prefer graceful fallbacks over surprise bills or silent failure modes.

For files and Spaces, we lean index-first: propose organization and relationships without moving your filesystem behind your back. That design choice costs us some demo “wow”—and earns trust over the long arc.

Shipping is iterative: tighter integrations, clearer permissions, and better retrieval beat a roadmap full of features nobody trusts enough to turn on.