Architecture Overview

ShipMyAgent is easiest to reason about with two top-level trunks:

• console: global control and operations plane
• agent: project runtime execution plane

Inside these trunks, service and extension form the capability layer.

Layered Structure

1. Console Layer (global)

• Commands: sma console ..., sma console config ..., sma service ..., sma extension ...
• Responsibility: global process management, operational entry points, model/runtime configuration governance

2. Agent Layer (project)

• Commands: sma agent create/on/off/status/restart ...
• Responsibility: load project context, run model/tool execution, persist runtime state

3. Capability Layer (runtime abilities)

• service: core domains (chat / skill / task / memory)
• extension: shared optional capabilities (for example voice)

4. State Layer (auditable persistence)

• Paths: .ship/context, .ship/logs, .ship/.cache, .ship/tasks, .ship/debug
• Goal: replayability, diagnosability, auditability

Architecture Rules

• Separate control from execution: console governs, agent executes.
• Separate core from optional: service is core, extension is enhancement.
• Keep command entry unified: same abilities are accessible through CLI/HTTP.
• Apply critical config first: model/runtime config is managed via console-facing commands before agent execution.

Read Next

• Architecture Logic Map
• Runtime Relationship & Process
• Service Runtime
• Extension Runtime