Complete autonomous AI agent memory and coordination system with:
- 4-layer memory architecture (Working, Session, Semantic, Knowledge Graph)
- Multi-agent coordination with deploy batching and resource claims
- MCP Router for 98%+ token reduction via hierarchical tool routing
- Rich session telemetry with cost tracking and pattern matching
- 22 behavioral patterns, 19 skills, 8 droids
- Harbor/Terminal-Bench integration for agent benchmarking
- Dashboard with 7 subcommands (overview, tasks, agents, memory, progress, stats, session)
- Session-start/pre-compact hooks for all coding harnesses
- Organized codebase: scripts/{benchmarks,setup,maintenance}, benchmarks/, docs/{8 categories}
- Centralized hooks via symlinks from templates/hooks/
- Full UAP naming standardization (no UAM references)
258 tests passing. TypeScript + Python. SQLite + Qdrant.