Concepts: Cicero and your office

Cicero comes in two layers. Knowing which one you are touching tells you what the wizard sets up and what stays yours.

Layer 1: Cicero

The product itself, identical for every user: the voice loop (hear, answer, speak, interrupt), the setup wizard, the office framework (employees, routing between them, memory) and sensible defaults. Cloning this repository and running setup gives you layer 1.

Layer 2: your office

Everything that is about you: which employees you have and how they talk, which agent profiles and models they run, and which hardware they run on. It lives in your ~/.cicero/config.yaml and never in the product's defaults, so nobody inherits someone else's office by installing Cicero. Packaging an office so it can be shared (an "office pack") is planned for a later release.

Default rule for employees

• Coding work goes to the agent's native harness: Claude Code or Codex, run the way their makers ship them.
• Personalities that should remember and learn over time go to Hermes, which can run any model.

What ships today

• Employees and routing exist as ACP office lanes (brain.lanes): several agents behind one call, transfers by voice, roll call. See the office.
• Employee templates and the office pack are planned (setup v2, part 2).
• Per-employee memory is planned (setup v2, part 3).
• The setup wizard sets up layer 1 plus one front desk and one agent. It does not create employees yet.

An example of layer 2

The project owner's own office (Hermes profiles, named characters, a reference GPU box) is documented as one example of layer 2, not as a default: see Advanced / example deployment.