Servers left running, copies filling your disk, models holding memory, plugins you never reviewed, and your usage across every account. One glance from your menu bar.
The real app. Pick a screen, or click the parrot in the menu bar.
Works with the agents and models you already use
Claude Code
OpenAI Codex
Gemini CLI
Cursor
Grok
OpenCode
Ollama
LM Studio
VS Code
Windsurf
Every server on your Mac, the agent behind it, and the ones still running after it quit.
Agents copy your project for every task. Pollymetric finds them all.
How much is left, when it resets, what's eating it, and which account to use next.
Every plugin and skill, what it runs, and what deserves a second look.
Hand any problem to your own agent. It explains first, and changes nothing without your okay.
No account, no tracking, no paid tier. Read every line, change it, share it. MIT licensed.
Open Pollymetric.dmg and drag it into Applications.
Start at login, full-disk checks, your AI agents. All optional.
It lives in your menu bar and speaks up only when something needs you.
A headless Mac mini's alerts, in your menu bar.
New packages, services and open ports an agent added.
Known-bad packages and pinning to versions.
Agent gateways and model servers reachable beyond your Mac.
Stop a process tree multiplying out of control.
Yes. Pollymetric is open source under the MIT license, with no account and no paid tier. It always will be.
No. Servers, worktrees, plugins, usage, history, cleanup and security checks all work on their own. An agent is only needed if you want one to explain or fix something.
Your monitoring history is stored on your Mac. Explain and Find a Fix share only the relevant context with the agent you choose. The app has no tracking. This website uses Google Analytics to count visits.
This early release isn't notarized by Apple yet. The first time you open it, go to System Settings → Privacy & Security and click Open Anyway. Notarized releases are on the way.
Claude Code, OpenAI Codex, Gemini CLI, Cursor, Grok and OpenCode, plus local models from Ollama, LM Studio, llama.cpp, MLX and vLLM, and plugins in Claude Desktop, VS Code and Windsurf.
It uses under 1% CPU at idle, and never runs heavy scans while you're busy.
Free, open source, and on your Mac in a minute.
