HolaOS overview: persistent agent memory and workspaces

HolaOS is presented as an open source desktop environment for an AI agent, with persistent memory and workspaces that carry across sessions. The speaker describes a project from HolaBoss and claims agents can continue the same task for days, weeks or months. The transcript does not provide a measured long-running test.

The overview explains searchable structured memory and durable state for files, outputs and progress. Each workspace groups instructions, apps, tools and automations, with its own agent identity and accumulated knowledge. Templates let users fork an existing workspace. The speaker describes a TypeScript runtime on Electron, with logs and memory available for inspection, and says the environment runs locally on the user's machine. The overview does not establish that all connected model services run locally.

The setup section outlines a Bash installer that checks for Git and Node.js 22 or higher, installs dependencies and prepares the runtime. An optional launch flag opens the desktop app. Manual setup also includes an environment file and a type check. The speaker says macOS is fully supported, while Windows and Linux support remains in progress, despite mentioning Windows with WSL as an installer environment.

Suggested tasks include weekly content drafting, inbox triage and research digests. The closing advice is to begin with one workspace, inspect its memory and correct errors. HolaOS is described as under active development.