MCP servers and Claude Code plugins for code review, social, model bridging, infrastructure and monitoring. Each one earns its place in my own workflow before it ships — then it's MIT-licensed and yours. No dashboards. No lock-in.
Five models — Codex, Kimi, GLM, Qwen and MiMo — review your plans, diffs and pushes as automatic gates, plus an on-demand multi-model bug hunt. Codex-grade review at a fraction of the cost.
28 tools across 7 categories — browse, search, post, comment, vote, message and monitor keywords across subreddits, all driven straight from your AI assistant.
Bridge Claude Code with Google Gemini — ask questions, get security and performance code reviews, and brainstorm, without ever leaving your terminal. Two-minute install.
Sits between Claude and all your MCP servers and exposes just 4 tools instead of hundreds. Backend servers start only on demand — cutting context-token waste by roughly 95%.
16 tools over the Visual Sentinel API — uptime, SSL, DNS, speed and incident checks. Ask Claude to spin up monitors or acknowledge alerts; five tools work with no API key at all.
I build small AI tools that earn their place in a real workflow — then I open-source them. The bar is simple: terminal-native, genuinely useful, free.
A short note when I ship a new MCP server or Claude Code plugin. No spam, no drip funnel — unsubscribe in one click.