Add new collection: AI Coding Agent Monitors

[tristan666666]

Adds a collection for a category that has grown quickly this year and has no collection yet: tools that watch AI coding agent sessions.

Disclosure first: one of the nine listed repositories, tristan666666/agent-island, is mine. I have ordered the list by how established each project actually is, which puts mine eighth. I would rather say that up front than have you find it.

Why this is a category and not a grab bag. Agents like Claude Code, Codex, Cursor and their peers write session records and expose usage endpoints, but none of them tells you the three things you actually need while working: how much quota is left, what the session is costing, and whether a run finished or stalled. A distinct tool class has formed around answering exactly that — ccusage alone is at ~17.8k stars — and it spans terminals, status bars and dashboards without overlapping the agent frameworks in collections/ai-agents, which are about building agents rather than watching them.

Selection. Every entry was checked for: pushed within the last 90 days, a real README, and a scope that is monitoring rather than something adjacent. Sorted by stars.

Repo	Stars	Form
ccusage/ccusage	17.8k	CLI, Rust
tddworks/ClaudeBar	1.4k	macOS menu bar, Swift
hoangsonww/Claude-Code-Agent-Monitor	894	dashboard, TypeScript
m1ckc3s/claude-status-bar	657	macOS menu bar, Swift
aqua5230/usage	287	macOS + Windows, Python
mag123c/toktrack	182	CLI, Rust
stefanprodan/cctop	130	terminal UI, TypeScript
tristan666666/agent-island	120	macOS notch + Windows tray, Swift/C#
minchenlee/c9watch	119	CLI, Rust

Deliberately excluded, so the reasoning is visible rather than implied: Dwtexe/cursor-stats (264 stars but no push since 2025-09-25); farion1231/cc-switch (126k stars, but it is a provider switcher, not a monitor — including it would have inflated the collection with something that does a different job); and several sub-20-star projects in the same space, for quality rather than scope reasons.

One collection only, per the contributing notes. No image included — collections/ai-agents ships without one, so I assume it is optional; happy to add one if you would prefer.

[github-actions]

Maintainer triage

Collection ai-coding-agent-monitors

Item	Stars	Last push	Owner type	Notes
ccusage/ccusage	17,830	2026-08-09	Organization	–
tddworks/ClaudeBar	1,400	2026-08-09	Organization	–
hoangsonww/Claude-Code-Agent-Monitor	894	2026-08-10	User	–
m1ckc3s/claude-status-bar	657	2026-08-09	User	–
aqua5230/usage	287	2026-08-10	User	–
mag123c/toktrack	182	2026-08-05	User	–
stefanprodan/cctop	130	2026-08-04	User	–
tristan666666/agent-island	120	2026-08-10	User	⚠️ possible self-submission
minchenlee/c9watch	119	2026-07-29	User	–

[tristan666666]

The ⚠️ possible self-submission flag on tristan666666/agent-island is correct — it is my project, and I disclosed it in the PR description before opening it. Flagging it is the right behaviour and I am glad the check exists.

Two things I want to put on the record so the flag does not have to be adjudicated blind:

It is placed where it actually ranks. The list is sorted by stars and mine sits eighth of nine, below ccusage (17.8k), ClaudeBar (1.4k), Claude-Code-Agent-Monitor (894), claude-status-bar (657), usage (287), toktrack (182) and cctop (130). I did not write myself to the top, and the triage table above confirms the ordering independently.

The collection stands without it. If you would rather not merge a collection containing the submitter's own repository, say so and I will remove tristan666666/agent-island and leave the other eight. The category is real and unrepresented either way — that is the point of the PR, not the entry. This is a genuine offer, not a rhetorical one; tell me and I will push the change.

One thing the triage table shows that argues for the collection independently of who filed it: every one of the nine was pushed within the last twelve days, and seven within the last week. This is an actively moving category with no collection covering it, adjacent to but distinct from collections/ai-agents — those are frameworks for building agents, these are tools for watching them.