auto agent

Automatically spawn and manage agents based on task requirements.

Usage

npx claude-flow auto agent [options]

Options

• --task, -t <description> - Task description for agent analysis
• --max-agents, -m <number> - Maximum agents to spawn (default: auto)
• --min-agents <number> - Minimum agents required (default: 1)
• --strategy, -s <type> - Selection strategy: optimal, minimal, balanced
• --no-spawn - Analyze only, don't spawn agents

Examples

Basic auto-spawning

npx claude-flow auto agent --task "Build a REST API with authentication"

Constrained spawning

npx claude-flow auto agent -t "Debug performance issue" --max-agents 3

Analysis only

npx claude-flow auto agent -t "Refactor codebase" --no-spawn

Minimal strategy

npx claude-flow auto agent -t "Fix bug in login" -s minimal

How It Works

1. Task Analysis

   • Parses task description
   • Identifies required skills
   • Estimates complexity
   • Determines parallelization opportunities

2. Agent Selection

   • Matches skills to agent types
   • Considers task dependencies
   • Optimizes for efficiency
   • Respects constraints

3. Topology Selection

   • Chooses optimal swarm structure
   • Configures communication patterns
   • Sets up coordination rules
   • Enables monitoring

4. Automatic Spawning

   • Creates selected agents
   • Assigns specific roles
   • Distributes subtasks
   • Initiates coordination

Agent Types Selected

• Architect: System design, architecture decisions
• Coder: Implementation, code generation
• Tester: Test creation, quality assurance
• Analyst: Performance, optimization
• Researcher: Documentation, best practices
• Coordinator: Task management, progress tracking

Strategies

Optimal

• Maximum efficiency
• May spawn more agents
• Best for complex tasks
• Highest resource usage

Minimal

• Minimum viable agents
• Conservative approach
• Good for simple tasks
• Lowest resource usage

Balanced

• Middle ground
• Adaptive to complexity
• Default strategy
• Good performance/resource ratio

Integration with Claude Code

// In Claude Code after auto-spawning
mcp__claude-flow__auto_agent {
  task: "Build authentication system",
  strategy: "balanced",
  maxAgents: 6
}

See Also

• agent spawn - Manual agent creation
• swarm init - Initialize swarm manually
• smart spawn - Intelligent agent spawning
• workflow select - Choose predefined workflows