Context-switching
is where work dies.
An AI-native operating system for the work you actually do — tasks, notes, habits, and the agents that carry them out.
§ 01Your agent already reads your code.
Give it your life.
The last two years taught us something strange: a programmer with a good CLI agent can do a week of work before lunch — but only on code. The moment the work touches a task list, a calendar, a relationship, a half-written note from last Tuesday, the agent goes blind. You paste context into chat. You translate. You context-switch.
Qalatra is the fix. A structured database your agent lives inside of — tasks, daily notes, habits, all machine-readable, all yours. No translation. No paste. No switch.
❝Not a task list your agent queries.
A system your agent works inside of.
§ 02Principles
- 01Local-first, always.Your tasks are a SQLite file. Your notes are markdown. If Qalatra disappeared tomorrow your data would still be yours, openable by any tool ever written.
- 02Agents for every mode of work.Some agents plan with you. Some just run. Some suggest — and that's exactly right. Qalatra doesn't dictate how your agents behave; it gives them the context to be useful either way.
- 03Dense and pretty.This is a tool for people who type faster than they click — and who still expect it to look good. Information density and good design aren't in conflict. We want both.
- 04Context lives across days.The most valuable thing your agent can have is continuity. Yesterday's note is today's system prompt. Last week's retro is next week's plan.
- 05Open, forever.MIT. Public roadmap. Every commit in public. If we break our own rules, fork us.
§ 03Why Qalatra?
Qalatra might be a word an AI made up.
We asked for a name and got back a beautiful backstory: a word from a Spanish-Arabic dialect meaning the collection of all knowledge — a repository of everything known. One reference online. Unvalidatable. Almost certainly a hallucination.
We kept it anyway. Not the etymology — the idea. The place where everything you know lives in a form your agent can actually use. Tasks, notes, habits, context, history. Whether the word is real or not, that's what we're building.
It felt like a fitting start for a tool built on the premise that AI will give you remarkable things — and you should think carefully about which ones to trust.