Agent Fleet Hud — AI skill for Claude Code
A notch-resident HUD for the AI coding agents already running on your Mac — Claude Code, Codex and Cursor: which is working, which needs an answer, which is rate-limited and when it resets.
How to install Agent Fleet Hud
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open valerijaspasojevic/agent-fleet-hud and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Agent Fleet Hud does
A notch-resident HUD for the AI coding agents already running on your Mac — Claude Code, Codex and Cursor: which is working, which needs an answer, which is rate-limited and when it resets. A monitor, not an agent framework.
Alternatives in AI
- CodeIsland — Real-time AI coding agent status panel in your MacBook notch — live status, approvals & replies for 13 AI tool 2.3k ★
- Wiki Query — Query the LLM Wiki and synthesize an answer 2.3k ★
- Claude Hub — by Claude Did This - A webhook service that connects Claude Code to GitHub repositories, enabling AI-powered c 393 ★
README
Agent Fleet HUD
A notch-resident HUD for the AI coding agents **already running** on your Mac.
It does not run, spawn or orchestrate agents. You start them however you normally do — `claude`, `codex`, Cursor — and this watches them: which is working, which is waiting on an answer, which is out of usage window and when it comes back. You can message one without leaving what you are doing.

*Rendered from the real panel with sample agents — `./make-screenshot.sh`.*
Collapsed, it's a small black tab hanging under the notch with live counts: red = blocked by the usage limit, blue = needs your input, green = working, amber = idle and waiting on you, slate = up but we cannot tell. It appears only while something is running and hides itself completely when nothing is.
Hover it to slide open the panel. Click the tab to pin it open (pinning also gives it keyboard focus so you can type a message).
What it detects
Nothing is hardcoded to one machine: apps are located by bundle identifier rather than assumed to be in `/Applications`, the build targets whatever architecture you build on, and every path is derived from your home directory.
| Source | How | Busy/idle | Send a message | Stop |
|---|---|---|---|---|
| Claude Code | ~/.claude/sessions/*.json |
yes, real status | yes, exact session | yes |
| Cursor | cursor-agent worker processes |
no — see below | yes, newest chat in the workspace | yes |
| Codex | the codex process — ChatGPT.app's app-server or the standalone CLI |
no — see below | yes, most recent thread | yes |
Messaging
Nothing here takes your focus. The message goes on the clipboard and the paste and return keys are posted straight to the target process with `CGEventPostToPid`, so the app never comes to the front — you keep working where you are while the message lands.
That needs Accessibility permission, because synthesising key events is what macOS gates. It does *not* need Au
Related Skills
Hermes Hud
Monitor AI agent memory, skills, and behavior in a live terminal HUD for Hermes.
Claude Notch Tracker
🦀 A Dynamic Island for your Mac's notch showing live Claude/Codex usage — 5-hour, 7-day, and credits with res
MioIsland
macOS Dynamic Island for AI coding agents. Monitor, approve, and jump to Claude Code sessions from the notch.
Agent Ping
When your AI agent needs an answer or approval, it pings you and waits. One-file Python CLI — works from Curso
Agent Fleet
Self-hosted web console for running AI coding agents (Claude Code, Codex, Copilot, Antigravity, Cursor, Kiro,
Agent Yes
Run AI coding agents (Claude, Codex, Gemini …) unattended — auto-answer prompts, auto-retry on rate limits, an
Related Agents
Pipeline Engineer
WS4 specialist for ingestion pipeline integrity in genai_incidents. Use for WS4-T1 through WS4-T8 - pinned ing
Buildhost Lead
Use PROACTIVELY and automatically — do not wait to be asked — as the Claude-Code node lead FOR an Apple-silico
Bonsai
Runs a task on the local Bonsai 2 27B model (PrismML ternary 27B, MLX, on this Mac) and returns its answer. Us