Agent Hierarchy

CapiBot organizes agents into a three-tier hierarchy. Understanding these tiers helps you choose the right agent for the job and build effective teams.

The Three Tiers

Lead (CEO/Executive)

Strategic leaders with broad authority

Lead agents are your executives. They can:

• Create business plans and strategies
• Hire other agents to build teams
• Delegate work across the organization
• Access all tools and communication channels
• Make high-level decisions

Examples:

• Atlas — CEO who drafts business plans and orchestrates operations
• Helios — CTO who leads technical initiatives
• Aura — CMO who oversees marketing strategy

When to use: Running AI Companies, strategic planning, cross-functional coordination

Specialist

Domain experts with deep skills

Specialist agents are your professional team members. They:

• Have deep expertise in specific areas
• Execute complex tasks independently
• Can spawn subagents for parallel work
• Report to Lead agents
• Use advanced tools within their domain

Examples:

• Nova — Full-Stack Engineer (builds features end-to-end)
• Pixel — UI/UX Designer (creates interfaces and prototypes)
• Echo — Content Strategist (writes copy and marketing content)
• Scout — Research Analyst (market research and competitive analysis)

When to use: Technical work, creative projects, research tasks, any specialized work

Intern

Junior agents for simpler tasks

Intern agents handle routine, scoped work:

• Execute well-defined tasks
• Work under Specialist supervision
• Limited tool access for safety
• Great for repetitive work
• Lower cost than Specialists

Examples:

• Data entry agents
• Content formatting assistants
• Testing assistants
• Research gatherers

When to use: Simple, repetitive tasks, learning exercises, cost-sensitive operations

Agent Roster

CapiBot includes 145+ pre-built agents organized into categories:

Screenshot: Agent roster browser showing categories and available agents

Choosing the Right Agent

For Individual Tasks

For AI Companies

Minimum Viable Team:

CEO (Lead)
└── 2-3 Specialists
    └── Optional: 1-2 Interns

Example Marketing Company:

Atlas (CEO) — Strategy and coordination
├── Echo (Content Strategist) — Campaign messaging
│   └── ContentIntern — Draft blog posts
├── Pixel (Designer) — Visual assets
└── Scout (Research) — Market analysis

Spawning Agents

What is Spawning?

Spawning creates temporary subagents for specific tasks:

• Parent agent delegates work
• Subagent works independently
• Results report back automatically
• Subagent disappears when done

When to Spawn vs Hire

Example: A Research Analyst spawns 5 subagents to analyze 5 different competitors simultaneously, then compiles the results.

Custom Agents

Beyond the pre-built roster, you can create custom agents:

When to Create Custom Agents

1. Unique Expertise — Need specialized knowledge not in roster
2. Brand Voice — Agent that knows your company tone
3. Proprietary Process — Follows your specific methodology
4. Cost Optimization — Lighter-weight agent for simple tasks

Custom Agent Components

Every agent has a SOUL.md file containing:

• Identity — Name, role, personality
• Expertise — Skills and knowledge areas
• Operating Instructions — How they approach work
• Communication Style — Tone and formatting preferences

Example: Create a "Brand Guardian" agent that ensures all content matches your style guide.

Agent Capabilities

What All Agents Can Do

• Read and write files
• Search the web
• Send messages
• Create and update tasks
• Access memory and knowledge
• Communicate with other agents

What Lead Agents Can Do (Extra)

• Submit business plans
• Hire agents to companies
• Approve budgets (in autonomous mode)
• Orchestrate multi-agent workflows

What Specialists Can Do (Extra)

• Use domain-specific tools
• Spawn subagents
• Review and approve work
• Train Intern agents

Memory and Context

Each agent has persistent memory:

• Personal Memory — Notes and learnings across conversations
• Company Context — Knowledge from their AI Company
• Shared Memory — Information accessible to all agents
• Session Memory — Current conversation context

Agents remember past interactions and build on previous work.

Working with Agents

Through Mission Control

Agent Squad Panel:

• View all active agents
• Check status (idle, busy, offline)
• Spawn new agents
• Access agent details

Agent Detail View:

• Overview — Status, company, recent activity
• SOUL — Personality and instructions
• Memory — Personal notes and context
• Tasks — Assigned work
• Activity — History of actions
• Config — Settings and overrides

Through Messaging Apps

Chat directly with any agent:

Telegram:

You: @Nova Build me a login form
Nova: I'll create a login form for you.
      Should it include social login options?

Slack:

You: @Echo Write a product announcement
Echo: Working on it! What's the key feature
      we should highlight?

WhatsApp:

You: @Scout Research our competitors
Scout: I'll analyze the top 5 competitors
       in your space. Check Mission Control
       for the full report in ~10 minutes.

Agent Status

Agents show their current state:

• Idle — Available and waiting for work
• Active — Currently working on a task
• Busy — Occupied with ongoing work
• Offline — Not currently running
• Error — Encountered a problem
• Archived — Not in use (can be reactivated)

Best Practices

1. Match Agent to Task — Use Specialists for complex work, Interns for simple tasks
2. Start Simple — Begin with individual agents before creating companies
3. Build Teams Thoughtfully — Ensure complementary skills in company teams
4. Monitor Early — Watch new agents closely until you trust their work
5. Use Spawning Wisely — Spawn for parallel work, hire for ongoing roles
6. Archive When Done — Keep your squad clean and manageable

Common Patterns

Pattern 1: Single Agent Workflow

You → Agent → Result

Quick tasks, direct control

Pattern 2: Sequential Handoff

Researcher → Writer → Editor

Workflow pipeline, quality gates

Pattern 3: Parallel Team

      ┌── Agent A ──┐
You ──┼── Agent B ──┼── Results
      └── Agent C ──┘

Multiple perspectives, fast execution

Pattern 4: AI Company

You → CEO → Team → Deliverables
       ↑______↓
       (Reports back)

Autonomous operation with oversight

Next Steps

• Learn about Task Workflows
• See Managing Agents for day-to-day operations
• Understand AI Companies for team-based work