MCP servers
Tool infrastructure for AI agents. A fleet of Model Context Protocol servers that give Claude Code, or any MCP client, capabilities past text: ears, a job pipeline, a video watcher, a calendar.
The throughline for all of these: my agent should be able to do the thing as a structured tool call. Analyze a track, find a job, watch a video, schedule a meeting. Not me running four CLIs and pasting output between them. Most default to Ollama and run on my own hardware.
Claud-Ear · audio intelligence
I have around 5,000 tracks and I wanted my agent to understand them the way I do. Not "a 3 minute MP3" but "melancholic D minor indie rock with a prominent bassline." Existing tools were shallow, cloud-only, GUI-only, or single-purpose. Claud-Ear does the whole pipeline as tool calls, plus an autonomous mode that grinds through the library overnight. GitHub ↗
JobHound · job outreach pipeline
Ashby, Greenhouse, and Lever all have public APIs that list every open role, and people still scroll job boards manually. JobHound scans the APIs, scores listings against your skills with a keyword heuristic, queues the good ones, and hands tailoring to Claude Code for per-role CVs and cover letters. SQLite pipeline tracking, a Textual TUI dashboard, a background scan daemon. I built it for my own hunt and the honest readme says where the autonomy stops: most boards still make you click submit yourself. GitHub ↗
Deep Video Watcher · video perception
I watch too many breakdowns and theory videos and forget 90% of them. This is a multi-modal perception engine, not a scene detector. It fuses vision, optical flow, audio stems, speech, and narrative analysis into one timeline: "bass drop, camera zoom, character scream at 12.3s." It can export highlight clips and an EDL for DaVinci Resolve, and learn editing style from a reference video. GitHub ↗
The small sharp ones
All open source under github.com/Null-Phnix.