Agentic OS

An agentic OS (agentic operating system) is an execution environment that gives AI agents the primitives a traditional operating system gives programs: identity, permissions, packages, a system-call surface, and logs. In an agentic OS, agents are first-class residents — each with its own identity, relationship-based permission scope, tool surface (such as MCP), and audit trail — rather than scripts behind a shared API key. ANGEE is an open-source, sovereign agentic OS.

The mapping is concrete, not metaphorical. In ANGEE: users and groups are the iam addon with Zanzibar-style REBAC permissions; processes are agent records with an owner and a lifecycle (draft, provisioned, ready, retired); packages are addons that install models, API, permissions, and UI deterministically; system calls are permission-scoped MCP tools; the system log is a replayable, attributed audit trail; and boot is one command — angee up.

An agentic OS is deliberately domain-blind — it does not know what an invoice or a heart-rate log is. Editions install a domain ontology on the OS: the typed entities, relationships, and permissions of a particular world, plus the agents that work it. fyltr installs the ontology of a personal life; ARPEE installs the ontology of a company. "Sovereign" describes where it runs and who controls it: AGPL open source on standard PostgreSQL, self-hosted on your infrastructure, with your choice of model providers.