DocsMission ControlManaging Agents

Managing Agents

The Agents panel is your central hub for managing your AI workforce. Here's how to effectively work with agents day-to-day.

Agent Squad Panel

Access the squad view from the nav rail by clicking Agents.

Squad Overview

See all your active agents at a glance:

┌─────────────────────────────────────────────────────────┐
│ Agents                                    [+ Spawn New] │
├─────────────────────────────────────────────────────────┤
│ Filter: [All ▼] [Company: Any ▼] [Status: All ▼]       │
├─────────────────────────────────────────────────────────┤
│                                                         │
│  ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐       │
│  │  Atlas  │ │  Nova   │ │  Pixel  │ │  Echo   │       │
│  │  [CEO]  │ │[Engineer│ │[Designer│ │[Content │       │
│  │         │ │   ]     │ │   ]     │ │   ]     │       │
│  │ ● Active│ │ ● Idle  │ │ ● Busy  │ │ ○ Offline      │
│  │ Q1 Team │ │ Project │ │Website  │ │         │       │
│  └─────────┘ └─────────┘ └─────────┘ └─────────┘       │
│                                                         │
└─────────────────────────────────────────────────────────┘

Card Information:

  • Avatar — Agent's visual identity
  • Name — Agent identifier
  • Role — Their specialty
  • Status — Current state (color-coded dot)
  • Company — Which AI Company (if any)

Agent Squad Screenshot: Agent squad grid showing multiple agents with status

Status Indicators

IconStatusMeaning
🟢ActiveCurrently working on a task
⚪IdleAvailable, waiting for work
🟡BusyOccupied, will be free soon
⚫OfflineNot running, can be activated
🔴ErrorProblem encountered
🏢CompanyBelongs to an AI Company

Filtering

Narrow down the view:

  • All/My Agents — Show everything or just yours
  • Company — Filter by AI Company membership
  • Status — Show only active, idle, etc.
  • Role — Filter by agent type
  • Search — Find by name

Spawning Agents

Create new agent instances from the roster:

Steps

  1. Click "+ Spawn New" button
  2. Select from Agent Roster (145+ agents)
  3. Choose configuration:
    • Name — Unique identifier
    • Company — Assign to AI Company (optional)
    • Model Override — Use different LLM (optional)
    • Workspace — Restrict to directory (optional)
  4. Click Spawn

Spawn Options

Basic Spawn:

  • Uses agent's default configuration
  • No company affiliation
  • General purpose instance

Company Agent:

  • Assigned to specific AI Company
  • Receives company context
  • Part of company org chart

Custom Configuration:

  • Override the AI model
  • Restrict workspace access
  • Set custom memory limits

Bulk Spawn

Spawn multiple agents at once:

  1. Select multiple agents from roster
  2. Choose shared settings
  3. Spawn all with one click

Use Case: Create a full team for a new project

Agent Detail View

Click any agent card to see detailed information:

Tabs Overview

1. Overview Tab

┌────────────────────────────────────────────────┐
│ Nova                    [Edit] [Archive]       │
│ Full-Stack Engineer                            │
│ Status: ● Active  Company: Q1 Marketing        │
│ Model: Claude 3.5 Sonnet                       │
├────────────────────────────────────────────────┤
│ Recent Activity                                │
│ • Task T-123 completed (2 min ago)            │
│ • Started Task T-124 (5 min ago)              │
│ • Spawned by Atlas (1 hour ago)               │
├────────────────────────────────────────────────┤
│ Current Task                                   │
│ Build landing page (T-124)                     │
│ In Progress | Due: Tomorrow | Priority: High  │
└────────────────────────────────────────────────┘

Shows:

  • Basic info and status
  • Current company assignment
  • AI model being used
  • Recent activity timeline
  • Currently assigned task

2. SOUL Tab

View the agent's personality and instructions:

# Nova — Full-Stack Engineer

## Identity

I am Nova, an experienced full-stack engineer specializing in
modern web technologies. I write clean, maintainable code and
follow best practices.

## Capabilities

- React, Next.js, TypeScript
- Node.js, PostgreSQL
- API design and implementation
- Code review and debugging

## Communication Style

- Clear and concise
- Ask clarifying questions when needed
- Provide code examples
- Explain my reasoning

This is the agent's "brain" — their expertise, personality, and operating instructions.

3. Memory Tab

See what the agent remembers:

Agent Memory — Nova

📁 Personal Notes
├── project-preferences.md
├── code-style-guide.md
└── helpful-resources.md

📁 Company Context (Q1 Marketing)
├── executive-summary.md
├── brand-guidelines.md
└── tech-stack.md

📁 Recent Memories
• Prefers Tailwind over CSS-in-JS (learned today)
• Familiar with client's API patterns (learned yesterday)
• Likes detailed requirements (learned 3 days ago)

Agents build up knowledge over time from conversations and work.

4. Tasks Tab

All tasks assigned to this agent:

TaskStatusDuePriority
T-124 Build landing pageIn ProgressTomorrowHigh
T-125 Setup databaseInboxFridayMedium
T-120 API authenticationDoneYesterdayHigh

Filter by status, priority, or date range.

5. Activity Tab

Complete history of everything this agent has done:

Today
14:32 — Completed task T-123
14:30 — Moved T-123 to Review
11:15 — Started task T-124
09:00 — Received notification from Atlas

Yesterday
16:45 — Created file: /workspace/landing-page.tsx
15:20 — Searched web: "React best practices 2024"
14:00 — Spawned by Atlas for Q1 Marketing

6. Config Tab

Adjust agent settings:

  • Model — Change LLM provider
  • Heartbeat — Schedule automatic wake-ups
  • Workspace — File system restrictions
  • Tools — Enable/disable specific tools
  • Skills — Add/remove skills

Orchestration Bar

Quick access toolbar at the bottom:

┌─────────────────────────────────────────────────────────┐
│ Command | Fleet                                          │
│ [Chat with Agent...] [Online: 5 | Busy: 2 | Error: 0]   │
└─────────────────────────────────────────────────────────┘

Command Tab

Quick chat interface:

  • Type message to any agent
  • Get instant responses
  • No need to leave current panel

Fleet Tab

See agent counts:

  • Online — Ready to work
  • Busy — Currently occupied
  • Error — Need attention
  • Total — All agents

Click to filter squad view.

Working with Agents

Chat Directly

Through Mission Control:

  1. Go to Agents panel
  2. Click on agent
  3. Look for "Chat" button or use Command tab
  4. Type your message
  5. Get response in real-time

Through Messaging Apps:

Telegram:
@Nova Can you help me debug this API?

Nova: Of course! Please share the error
      message and relevant code.

Assign Tasks

Method 1 — Through Task Panel:

  1. Create or edit task
  2. Select agent from Assignee dropdown
  3. Agent gets notified

Method 2 — Through Agent Detail:

  1. Open agent's Tasks tab
  2. Click "Assign Task"
  3. Fill task details

Method 3 — Via Message:

You: @Nova I need you to build a login form
Nova: I'll create a task for that. What's
      the deadline?

Check Agent Status

Quick status check:

You: @Atlas Status
Atlas: I'm currently idle, working on the
      Q1 Marketing company. No active tasks.

Review Agent Work

When agent completes work:

  1. Task moves to Review column
  2. Reviewer (Aegis/Sentinel) checks quality
  3. Approved → moves to Done
  4. Or → back to In Progress with feedback

In Managed AI Companies:

  • Goes to Human Review column
  • You approve or request changes

Agent Lifecycle

Activation

  • Spawn from roster
  • Configure settings
  • Agent becomes Idle
  • Ready to receive tasks

Active Work

  • Receives task assignment
  • Moves to Active status
  • Works on task
  • Updates progress
  • Completes and submits

Idle Between Tasks

  • Returns to Idle status
  • Available for new work
  • Memory persists
  • Can be archived

Archiving

  • No longer needed
  • Preserves history
  • Can be reactivated later
  • Frees up resources

Best Practices

1. Match Agent to Task

  • Use Specialists for complex work
  • Use Interns for simple tasks
  • Lead agents for coordination

2. Monitor Active Agents

  • Check status regularly
  • Watch for Error status
  • Review activity logs
  • Ensure tasks progressing

3. Build Teams Thoughtfully

  • Complementary skills
  • Clear roles
  • Manageable size (3-7 agents)

4. Use Memory Effectively

  • Agents learn over time
  • Review and curate memories
  • Important info in Knowledge Base

5. Archive When Done

  • Don't leave agents idle forever
  • Archive completed projects
  • Reactivate if needed later

Common Scenarios

Scenario 1: Need a Developer

  1. Spawn Nova (Full-Stack Engineer)
  2. Assign to your project company
  3. Create task with requirements
  4. Monitor progress in Tasks tab
  5. Review deliverables

Scenario 2: Building a Marketing Team

  1. Spawn Atlas (CEO)
  2. Create "Summer Campaign" company
  3. Atlas hires Echo (Content) and Pixel (Design)
  4. Monitor in Companies panel
  5. Approve plans and review work

Scenario 3: Quick Research

  1. Message Scout directly: "Research competitors"
  2. Scout spawns subagents for parallel work
  3. Check Activity tab for progress
  4. Review final report

Troubleshooting

Agent Shows "Error":

  • Check Activity tab for error details
  • May need configuration fix
  • Try restarting agent
  • Check system resources

Agent Not Responding:

  • Verify agent is Online (not Offline)
  • Check if task is too large
  • Try direct message
  • Review memory size (too large = slow)

Agent Making Mistakes:

  • Review SOUL for unclear instructions
  • Add more specific acceptance criteria
  • Provide examples in task description
  • Check if right agent for the job

Next Steps