Built, and planned

This page separates what works today from what is planned. Planned items carry no dates.

Built

Three engines: Claude Code, Codex and the native engine.

  • You decide what it may do
  • Projects are folders
  • Usage in plain sight
  • Rules it follows
  • A pet that shows the work
  • A graph you can look at
  • Prompt queue
  • Slash commands
  • Images
  • Models and effort
  • Keys
  • Clean exit

Planned

  • A learning loop: memory, skills and session search offered to the engines. Parts of the brain are already in the app.
  • Theming: a settings panel, a background image, hot reload, import and export.
  • Scheduled jobs that run unattended, with a tray keep-alive and notifications.
  • Plugins: tools, commands, hooks and external MCP servers.
  • Installers for Windows, macOS and Linux, with signed auto-update.
  • Hosted provider adapters for the native engine.
  • Agent teams: a tree of agents, each with its own name, role, instructions, engine, model and permission mode.
  • A built-in browser panel and browser tools, driving a Chromium-based browser you installed, with a separate profile.
  • Computer use: the agent reads and acts on other desktop programs. Off by default, with a permission card for each action.
  • A quiet notice when Aurelius learns a skill, with View and Undo.
  • A pop-out pet in its own always-on-top window.

Not in version 1

  • A messaging gateway.
  • Editing files from the file tree.
Back to the overview