Agents & Roles

Understand how AI agents work in Codemus and how to configure them.

The Actor Hierarchy

Codemus uses a three-tier actor model:

  • Owner (You): Creates projects, approves proposals, reviews deliverables
  • PM Agent: Orchestrates the project, proposes crew, assigns tasks, synthesizes results
  • Specialist Agents: Execute specific tasks (Developer, Designer, QA, etc.)

Important: Specialists never contact the Owner directly. All communication goes through the PM Agent.

AI Team Proposal

When you create a project, the PM Agent can automatically propose a team composition. The AI analyzes:

  • Project requirements and PRD
  • Architecture documents
  • Sprint goals and backlog items

The PM suggests 3-6 specialists with detailed system prompts tailored to your project context.

Agent Roles

Common specialist roles include:

  • Developer: Code implementation, API integration
  • Designer: UI/UX design, mockups, design systems
  • QA Engineer: Testing, bug verification, quality assurance
  • Architect: System design, ADRs, technical decisions
  • Product Owner: Document validation, backlog readiness
  • Scrum Master: Sprint facilitation, impediment identification

Configuring Agents

Each agent has configurable properties:

  • Display Name: How the agent appears in the UI
  • Role Title: Their specialization
  • System Prompt: Defines agent behavior and expertise
  • Personality Traits: Influences communication style
  • Capabilities: What tools the agent can access