What's New in v0.2.0

Demo Mode (--demo)

See Claude Swarm in action instantly — no API key needed! An animated TUI simulation showcases the full swarm experience with realistic task decomposition, parallel agent execution, and live dashboard updates. Perfect for presentations and hackathon judging.

claude-swarm --demo

Opus 4.6 Quality Gate (Phase 2.5)

After all agents complete their work, Opus 4.6 reviews the combined output for:

• Correctness — bugs, logic errors, security issues
• Consistency — do agents' outputs work together cohesively?
• Completeness — was the original task fully addressed?
• Integration — problems between different agents' work

This doubles the Opus 4.6 touchpoints (planning + review), mirroring how senior engineers review junior engineers' parallel work.

Task Retry

Failed tasks are automatically retried with configurable limits:

claude-swarm --retry 3 "Build auth system"  # up to 3 attempts per task

Stats

• 44 tests passing (30 → 44)
• 10 source modules (8 → 10)
• 2,693 total lines (1,488 → 2,693)
• 12 CLI flags (8 → 12)

All Changes

• --demo flag for animated TUI simulation (2 built-in scenarios)
• --quality-gate/--no-quality-gate to enable/disable Opus review
• --retry flag for configurable task retry limits
• Quality report display in UI (score, verdict, per-task reviews)
• Quality gate events in session replay
• Architecture diagram updated with Phase 2.5
• Comprehensive "How Opus 4.6 Is Used" documentation