I build tooling for AI coding agents — making multi-agent workflows safer, more reproducible, and easier to orchestrate.
| Project | What it does |
|---|---|
| mantra | Repo-ops harness for Claude Code — bounded continuous improvement via autonomous-improvement-loop, specialist agents, and session continuity |
| maw | Multi-Agent Workspace manager — lightweight parallel workspaces for AI coding agents using git worktrees |
| myak | Persistent memory for Claude Code — auto-index session transcripts into SQLite FTS5, searchable across projects and sessions |
| kxxx | Agent-safe secret runtime — broker operations through the CLI so LLMs never see raw credentials |
| ma-skills | UI design & review skills for AI coding agents, grounded in the philosophy of 間 (ma) |
AI Agents Developer Tools Claude Code Multi-Agent Systems Design Systems



