Quick answer
To check memory pressure and storage before you buy a new computer, run your normal AI agent work and watch the memory pressure graph in Activity Monitor (on Windows, the Memory view in Task Manager). Green means you are fine, yellow or red means strain. Then check storage, list what is still running from finished jobs, and ask your agent for a read-only report of the biggest memory and disk users. Close or clear what you no longer need and test again. If you are still in the red with only useful work running, you have real evidence that you need more.
Key takeaways
- Test while your normal agent workload is running, not on a quiet machine
- On a Mac, memory pressure colour matters more than the “memory used” number
- Look for leftover dev servers, browsers and chats from finished jobs
- Ask your agent for a read-only audit, and approve each clean-up yourself
- Re-test after clearing, so you compare before and after
- Only buy when the machine still struggles with work you actually need
📘 Main article: This guide supports Computer for AI Agents: How Much Memory and Storage You Need. It turns that article’s “audit first” advice into a step-by-step check you can run today.
A slow computer feels like a hardware problem. Often it is a tidiness problem, especially if AI agents have been busy on it.
This guide answers one question: how do you check memory pressure and storage before you buy a new computer? It supports our main article, Computer for AI Agents: How Much Memory and Storage You Need, based on Nate B Jones’s video on agent computers.
Why check memory pressure before buying?
Because more memory does not fix agents that never stop. Nate’s point is that you can buy more capacity, but you cannot buy your way out of the habit. A 20-minute check tells you whether you need new hardware or just a clear-out.
How do you check memory pressure on a Mac?
- Start your usual agent work: the coding agent, browser, chat apps and anything else you normally run.
- Open Activity Monitor (search for it with Spotlight) and choose the Memory tab.
- Watch the memory pressure graph at the bottom for 10 to 15 minutes.
- Sort by Memory to see which apps use the most.
- Note the colour: green, yellow or red, and when it changes.
| Colour | Meaning | Action |
|---|---|---|
| Green | Coping well | No hardware change needed for this workload |
| Yellow | Under strain | Close leftovers, then re-test |
| Red | Struggling and swapping heavily | Audit carefully; if it stays red with only useful work, consider more memory |
💡 Tip: A high “memory used” figure on its own is not a problem. macOS uses spare memory on purpose. The pressure colour is the better signal.
How do you check memory use on Windows?
Open Task Manager, go to Performance, then Memory. Watch how much is in use and the committed figure while your agents run, and use the Processes tab to sort by memory. If you are regularly near the limit with only useful work running, that is a sign you need more.
How do you check storage?
On a Mac, open System Settings, General, Storage. On Windows, open Settings, System, Storage. Look for:
- Free space: very low free space slows swap and updates
- Developer files and old project folders with build output
- Large log or crash folders from AI tools (Nate cites a public Codex report of one over 50GB)
- Old local models you downloaded and no longer use
⚠️ Warning: Do not delete folders just because they are large. Find out what each belongs to first, and keep anything that holds active work.
How can your agent run the audit for you?
Ask for a report, not a clean-up. Agents are good at this because they can read system information quickly. A prompt in the spirit of Nate’s:
“Inspect my computer but do not change anything. List the biggest memory users and the largest folders, explain what each belongs to, flag anything left over from finished work, and suggest specific fixes. Wait for my approval before doing anything.”
Then approve fixes one at a time: stop a finished dev server, close an old browser session, clear a build folder you can regenerate. Our Claude and Codex setup audit pairs well with this.
How do you read the results?
| After clearing leftovers… | What it suggests |
|---|---|
| Pressure drops to green | You had a tidiness problem, not a hardware one. Add a weekly check-up. |
| Yellow only during big jobs | Queue heavy jobs and run fewer at once. |
| Still red with useful work only | You have evidence for more memory. See our comparison of machines. |
| Storage nearly full | Clear regenerable files or add storage; consider 1TB if you work with video. |
Need more machine after all? Compare your options in Mac, NVIDIA or Your Old Computer: Which Should Run Your AI Agents?.
Frequently asked questions
What is memory pressure on a Mac?
It is a graph in Activity Monitor that shows how hard macOS is working to manage memory. Green is fine, yellow is strain and red means the machine is struggling.
Is high RAM usage bad?
Not by itself. Operating systems use spare memory deliberately. Pressure and slowdowns are the signs to watch.
Can I let my AI agent clean up my computer?
Yes, but start with a read-only report and approve each change yourself.
How often should I run this check?
Weekly is a good habit if agents run on your machine every day, plus whenever it feels slow.
How much memory should a new computer for AI agents have?
Nate B Jones suggests 24 to 32GB for most agent users. See Computer for AI Agents: How Much Memory and Storage You Need.
The bottom line
Before you buy, check memory pressure and storage while your agents do real work, clear what is left over from finished jobs, and test again. If the machine still struggles with only useful work running, you have a solid reason to upgrade. If not, you have saved yourself the cost of a new computer.
Related posts
- How to Audit Your Claude and Codex Setup and Clear Out AI Clutter
- How to Run Long AI Agent Jobs Overnight Without Wasting Money
- How to Split a Task Between Parallel AI Agents Without Making a Mess
