I build tools for AI coding agents and ship them across many agent hosts at once. The M365 security work is the domain I apply inside that frame.
→ CloudJourneyBlog · LinkedIn · X
OpenLtm — long-term memory for Claude Code. Semantic search over what I've already worked out, context injected back into the session, learning so the next session starts further along. The only repo I work on daily. TypeScript, MIT.
microsoft-graph-security (private) — M365 security investigation skills, packaged once for many agents, with 47 documented hosts. Read-only by default; writes need explicit opt-in.
Omarchy bar widgets — a QML desktop stack for observing what agents actually do: agent-monitor · docker-monitor · rohi.network · plugin-lab
skills — my everyday agent skills library, for Claude Code, opencode and friends.
argus — self-hosted, Dockerized notification and reporting for M365, for IT admins and security teams. My most-starred repo and the one I have not touched since June; MIT is on the way in argus#7.
Also: fleetwatch (Intune inventory) · synapse-mcp (tenant-isolated MCP gateway) · aegis-security-agent · corporate-on-demand
Contributing outside my own repos: NousResearch/hermes-agent, omacom/omarchy and the plugin marketplace. Most of what I file against my own estate is security findings, not features.
Stack — TypeScript · Python · Next.js · PostgreSQL · Docker · M365 · Azure · GCP · Claude Code · MCP · QML · PowerShell
Powered by caffeine and questionable life choices.