I built this because I was running Claude Code, a local LLM, and multiple dev servers simultaneously and had no visibility into what was actually happening. Activity Monitor is useless for this. htop has no context.

Hydra is a macOS desktop app (Electron + React + TypeScript) that shows:

- Which AI agents are running (Claude Code, Codex, Cursor, Gemini) and their status - Per-process CPU/memory with project grouping - Port-to-process mapping - Git repo health across all your projects - Network bandwidth per process - Security posture via a built-in scanner - AI briefings from a local LLM (I use LM Studio with Qwen 35B) — no cloud API needed - Claude Code usage and cost tracking

Everything runs locally. No telemetry, no accounts, no cloud dependency.

GitHub: https://github.com/kunalnano/hydra

Happy to answer questions about the architecture or the local LLM integration approach.