Vitals now has an MCP server for Claude, Codex, and Cursor
AI assistants already work on my Mac. They start dev servers, run builds, and when something is slow, they try to find out why.
All they see is a thousand process names, so they guess. And when they clean up, they can't tell a dev server I forgot last week from the database I'm working in.
So Vitals now answers them.
Ask: what is slowing the Mac down, which app is behind a number, what is on port 3000, which dev servers have sat idle for days.
Quit: an assistant can't quit anything by itself. Vitals shows you the request in its own window, and nothing is quit until you say so.
Share: it's off until you turn it on, and you choose what assistants may see.
Setup is one page in Settings. Turn it on, click Add beside Claude, Codex, or Cursor, and that's it.
On my own Mac, the first thing it found was a dev server that had been running for 16 hours and holding 3.8 GB. I had no idea it was still there.
If you let an assistant work on your Mac, what do you wish it knew before it touches anything?
The update is free for everyone who owns Vitals. Open Vitals and it will offer the update, or download it from vitalsmac.com.
How to use it, with prompts to try: vitalsmac.com/changelog/1.4
Replies
Whether another task still depends on that process
A quiet dev server may still be needed by a browser test
Would be useful to show those connections in the quit request so I can see what else might stop working
@al_kub Part of that is there today. Vitals checks whether anything is connected to a dev server right now, and if something is, it stops calling the quit safe and tells you so in the window.
What it doesn't do yet is name what is connected, like "Chrome" or "a test run in this project". That is the useful part of your suggestion, and I've added it to my list.
I see what you mean but a quiet dev server is rarely really idle. I've had browser tests depending on one without noticing, so naming the dependent task before quitting would save some pain. Is that in the plan?
@matheus_paranhos1 It shipped today, in 1.4.2. When something is connected to a dev server, the quit window now names it under "Connected now", so you see what else would stop before you say yes.
One limit: a browser driven by a test shows up as the browser, not as the test. A test runner or another project's process is named with its project.
Honestly, I think context is the missing piece. An agent can have access to a lot of tools, but without knowing what’s safe to touch, what belongs to the current project, and what state the system is in, it’s still making decisions with incomplete context
@nihal_pandey4 yeah that's why i built the MCP in Vitals. it tells the assistant which project a process belongs to, if anything's using it, and if quitting it will lose your work. if it still wants to kill something, you approve it, not the assistant.
what do agents get wrong most on your Mac?
@attacomsian For me, it’s mostly project intelligence. Agents don’t always know what project they’re actually operating on or what’s already running.
Then there’s context drift in long chats — they start assuming things instead of checking the actual state. And yeah, sometimes they take setup/actions without asking first