What is AI-POS?
AI-POS (AI Project Operating System) is a desktop app that organises any long-running project (writing software, doing research, running a side business, planning your life) into a standard structure that an AI can take over directly:
- Project (workspace): a folder you choose. All state lives inside it as human-readable files. There is no black-box database: you can inspect everything with your file manager or with git, any time.
- Tasks: atomic units of work, each small enough to finish within a single working session, with a status (todo → in progress → done, or blocked).
- Phases and gates (workflow): the project moves through phases defined by a template (for example plan → execute → review). Each phase has exit gates, and the engine decides whether they pass by checking real state (for example “all tasks are done”). You don’t get to mark your own homework.
- Handoff snapshot: a summary of “where things stand and what comes next”, maintained automatically by the engine. Switch to a new AI session, a different machine, or a different person: reading this one file is enough to pick up the thread.
In one sentence: you make the decisions, the AI does the work, and AI-POS makes sure a handoff is always possible.
Why this exists
Section titled “Why this exists”An AI coding session is short and forgetful; a real project is long and full of context. Most of the friction in working with AI agents comes from that mismatch: every new session starts by re-deriving what the last one already knew, and the parts that never got written down get lost.
AI-POS puts that context in files the engine keeps fresh, so the next session starts from state rather than from scratch.
What it is not
Section titled “What it is not”- Not an AI agent. AI-POS doesn’t write your code or your report. It is the management hub; the execution is done by whichever agent you prefer (Claude Code, Codex, Gemini CLI, or you).
- Not a cloud service. Your project state is local files on your machine, and AI-POS itself has no backend to send them to. It does not, however, make any claim about the agent CLIs you run inside it; those talk to their own vendors, with their own credentials, exactly as they would in an external terminal.
- Not a task tracker you have to feed by hand. The handoff snapshot regenerates on every state change, so it is never staler than your last action.
Where to go next
Section titled “Where to go next”- Install: get it running.
- Your first project: the initialisation wizard, start to finish.
- Connect an AI agent: the core scenario.