Non-Interactive Claude Flow Commands

Working Commands (alpha.83)

These commands work correctly with --non-interactive flag in claude-flow alpha.83:

1. Swarm Command

bash

# Basic swarm with non-interactive mode
npx [email protected] swarm "Your task description" --non-interactive

# With additional options
npx [email protected] swarm "Build a REST API" \
  --strategy development \
  --mode distributed \
  --max-agents 8 \
  --non-interactive

# With executor (for built-in execution without Claude CLI)
npx [email protected] swarm "Analyze code patterns" \
  --executor \
  --non-interactive

2. Hive-Mind Command

bash

# Basic hive-mind spawn
npx [email protected] hive-mind spawn "Design architecture" \
  --count 4 \
  --coordination majority \
  --non-interactive

# With different coordination modes
npx [email protected] hive-mind spawn "Solve complex problem" \
  --count 8 \
  --coordination weighted \
  --non-interactive

# Byzantine fault tolerant mode
npx [email protected] hive-mind spawn "Critical system design" \
  --count 10 \
  --coordination byzantine \
  --non-interactive

3. SPARC Commands

bash

# SPARC modes (these work without --non-interactive as they're already non-interactive)
npx [email protected] sparc spec "Define API requirements"
npx [email protected] sparc architect "Design microservices system"
npx [email protected] sparc tdd "Create authentication module"
npx [email protected] sparc integration "Integrate payment gateway"
npx [email protected] sparc refactor "Optimize database queries"

# With output format
npx [email protected] sparc tdd "Build calculator" --format markdown
npx [email protected] sparc architect "Design system" --format json

# With file output
npx [email protected] sparc tdd "Create parser" --file ./output/parser.md

Benchmark Usage

Using Real Execution with Benchmark Tool

bash

# Swarm benchmark
./benchmark/bin/swarm-benchmark real swarm "Your task" --strategy auto --mode centralized

# Hive-mind benchmark  
./benchmark/bin/swarm-benchmark real hive-mind "Your task" --max-workers 8 --consensus majority

# SPARC benchmark
./benchmark/bin/swarm-benchmark real sparc tdd "Your task"

Important Notes

Version Requirement: Use [email protected] for working non-interactive mode
Current Issue: Versions after alpha.83 (including alpha.87) have a regression where --non-interactive flag is not properly handled
SPARC Commands: Work without Claude CLI and don't require --non-interactive flag
Swarm/Hive-mind: Require Claude CLI unless using --executor flag

Troubleshooting

If Non-Interactive Mode Fails

Check claude-flow version: npx claude-flow --version
Use alpha.83: npx [email protected]
For SPARC modes, omit --non-interactive flag as they're already non-interactive
Use --executor flag for swarm/hive-mind to avoid Claude CLI dependency

Exit Codes

0: Success
-9: Timeout (SIGKILL)
1: General error
127: Command not found (usually Claude CLI missing)

Examples for Testing

bash

# Test swarm non-interactive
npx [email protected] swarm "Hello world test" --non-interactive

# Test hive-mind non-interactive
npx [email protected] hive-mind spawn "Simple task" --count 2 --non-interactive

# Test SPARC (already non-interactive)
npx [email protected] sparc tdd "Simple function"