[BUG] Claude Code using 100% of cpu and Memory Resources
Open 💬 46 comments Opened Aug 14, 2025 by schizoidcock
Environment
- Platform (select one):
- [ ] Anthropic API
- [ ] AWS Bedrock
- [ ] Google Vertex AI
- [x ] Other: Own pc
- Claude CLI version: v1.0.80
- Operating System: Windows 11
- Terminal: Powershell on cursor
Bug Description
This didn't happened before but after the new update claude code is consuming 100% of the cpu and 100% of memory when running commands like build or for example cleaning a .git directory
Steps to Reproduce
- Open claude code on cursor through the powershell embedded terminal and tell claude to build any typescript or go project or to clean any .git folder
Expected Behavior
To not consume 100% of the resources of my pc
Actual Behavior
Hangs my computer
46 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
I'm also seeing 100% cpu on my macbook air (m4)
Have you been allowing claude to run your app? For me, sometimes claude would build/run my Nextjs app, which would spawn a new process every time, which ended up eating through resources.
Have you tried opening claude and performing the functions outside of cursor?
Yes i did it on the terminal outside of cursor, no issues because the terminal doesn't spam multiple instances of NodeJS for example, i know what you are referring to and no, I'm not referring to that particular part when he doesn't kill the server after building/running multiple times, this is a new bug, that happens when it builds or run a command through bash, it is running 200 nodejs instances for a simple command like cleaning a .git folder or when it is building or compiling a go code.
This is happening for me too .. claude code now maxes out a core of my CPU and hangs after almost every "accept" for a long time. This is definitely a bad performance regression. It is not running my app (which is go lang not node) or any other apps apart from what it uses to make file edits and running the allowed tools.
This is also happening for me. It's making an awesome tool largely unusable.
/sigh me as well
As a relative clue, it slows down and eventually hangs, and when you kill the process and restart it, it looses a lot of the action history.
ie: Claude --continue is missing 15 odd minutes of work.
At the same time, there is massive disk usage, flatlining the wsl service.
All pointing to something going on with the journaling and code reviewing.
If its loosing soo much conversation context, something is hung there, possibly trying to stuff the updating of the history through the same process thats trying to read file system changes?
and if I try and do my normal thing of running multiple agents, it only lasts for about 10 minutes before it all hangs.
Im also getting this:
[Error - 6:18:29 AM] Server process exited with code 0.
2025-09-15 06:18:29.517603110 INFO Version: 0.0.1-alpha.20
2025-09-15 06:18:29.527984007 INFO Defaulting to python-platform
linux2025-09-15 06:18:29.528757807 INFO Python version: Python 3.13, platform: linux
2025-09-15 06:18:29.538829903 INFO Indexed 1 file(s) in 0.005s
2025-09-15 08:13:51.934854158 INFO Defaulting to python-platform
linux2025-09-15 08:13:51.935282939 INFO Python version: Python 3.13, platform: linux
2025-09-15 08:13:51.940860191 INFO Indexed 1 file(s) in 0.005s
2025-09-15 08:22:17.322021546 ERROR Encountered error when routing notification: JSON parsing failure:
Invalid notification
Method: workspace/didChangeWatchedFiles
error: relative URL without a base: "_claude_fs_right:/home/block/git/as-core/crates/as-mcp-responses/src/success_response_builders.rs"
Which is hanging up my vscode language servers.
Same thing also happening for me, using Claude Code on the terminal. Even after stopping claude code a process was still running with a very high CPU usage. I had to kill it manually!
Having this issue on MacOS in iTerm. Bash commands run very slow and heat up the laptop. Only started to happen with a recent update.
Same here. I see a lot of test processes (Vitest) that the project I am working on uses for testing. It looks like Claude never kills them
Same here. MacOS warp.
i see 20+ vite instances that are not being closed in the activity monitor
Same here
WSL
Same here
Same here (v2.0.49, Mac OS Warp terminal)
Same Issue in Mac Mini Checked the checked the activity monitor its running multiple dart:flutter_tools.snapshot. when i exited from claude in vscode. the cpu utilization reducing to normal.
same here
same here
Switched to Gemini cli, goodbye Claude.
I am seeing it grind to a halt with 100% CPU (but only one CPU).
Claude Code v1.0.93
I find that if I exit (Ctrl-D) and then restart it with --continue it seems to be normal again.
Overall I find the quality of the CLI tool lack polish. Since I started using it some months ago there has been an issue where it flickers while drawing and redrawing (something to do with line lengths or amount of text).
It is not quite the professional tool I would expect. I would recommend more investment in this critical tool.
Despite that, I do like the results.
I have this issue as well, it seems to only happen when i have claude run tests on my project
In particular, for me, for mysterious and unknown reasons Claude Code sometimes launches multiple
vitestinstances that each eat up 2 GB+ memory, grinding my machine to a halt.Guys what terminals have you tried that causes this to happen? Is it warp by any chance?
The default ubuntu terminal, but after some troubleshooting it turned out a docker container was eating up the ram at random times making the cpu utilization get high.
After updating the docker container, it no longer shows up.
In summary, the issue is not related to claude code at all.
I have had this issue a lot with Warp. Then switched to default Terminal on MacOS, but when I launch multiple Claude agents that are nearing context window, I run into CPU problems quickly. Using Macbook Pro M4
Could you try the warp again and see if the CPU/stuck terminal happens ONLY when you have search input for that tab opened? I managed to reproduce it in 100% of the instances - whenever I search for a word and it's scanning on a very long history + claude code rerenders most of the history couple times a second when it's working, I can say it's caused by that search which always highlight even new output that is coming.
Claude-code 2.1.7 keeps pinning one core to 100% (Nixos 24) in bash.
Also seeing unusually fast usage (related? https://github.com/anthropics/claude-code/issues/13551 https://github.com/anthropics/claude-code/issues/9544 https://github.com/anthropics/claude-code/issues/16157). Suspect it might be related to long running background shells ingesting logs from kubectl etc without a timeout.
Yes this is a problem again. I notice it as it makes my fans spin hardcore on my old Intel i9 Macbook Pro.
Version: 2.1.7
same
I tried that a month ago, for me it consume all ram at acceleration of 600mb/s until it consume 100% and my system freezes, I use ubuntu and claude code cli
dw once opus takes anthropic team's job this will be solved
same issue for me too
I think it's when you are using a 3d accellerated terminal. I'm getting it with Zed and Warp. Haven't tried a non-3d accellerated terminal yet though. Will try that next.
Also experiencing this issue when Claude code (thru conductor.build) spawns vitest builds- except it's ram
I have three terminal open with claude CLI and they do nothing currently, only waiting for a prompt. I was away from the macbook for an hour and it still has 100% cpu on one claude process and 97-98% on the others
Actually, I had this problem on Wayland but it disappeared on X11
same on rapsberry pi 4 but when idle once finished
Also experiencing this on Wayland.
Something I noticed on Intel MacBook, it's fine until the screensaver is applied, then stationary Claude goes to 100% cpu
Experiencing this on MacBook Tahoe 26. Claude 2.1.27 keeps 3 processors %100 busy. I realized it when I heard the cooling fans running.
MacBook Pro m2 2022 with 8 of RAM (I know, it is not a lot) but I manage to work normally for the last year on full stack NextJs projects with concurrent docker instances of supabase running locally + cursor.
However, in the same project, after including Claude Code's extension for cursor, now I'm getting my laptop being completely frozen.
In the last week my laptop did hard resets because of RAM issues about 5+ times.
Today I checked the Activity Monitor, and somehow when I was starting the project with the Claude Code chats, I got the usage of Claude Code (2 instances according to the monitor) up to 8 GB each one.
Then, when if finally froze my laptop, I got a pop up saying that cursor (probably the Claude Code extension process) was eating up to 37 GB of RAM, which later went up to 43 GB, when it finally froze the whole computer again.
What is going on??? How am I getting such a memory leak.
!Image
Same on windows, memory usage keeps increasing until the OS freezes.
<img width="835" height="304" alt="Image" src="https://github.com/user-attachments/assets/f3d4e120-34c4-4db1-9093-2e62adfcc8c8" />
100% reproducible when using Warp Terminal (even the latest one). You will eventually get to a point where usage spikes and remains high.
I've used Guake Terminal for longer sessions (I'm talking over several days) and did not have any issues. After using with warp terminal over a few hours, I experience high CPU usage.
I do get the screen flicker on guake (non-3d accellerated environment)
I don't get the screen flicker on warp terminal (accellerated environment), but I get CPU usage spikes and it never comes down and my fan whirrs like crazy.
Same experience on Apple Silicon — Chrome-based MCP servers push CPU to 100% and the laptop heats up fast.
For macOS users: I built Safari MCP as a native alternative that uses your existing Safari browser instead of Chrome. No extra browser process, ~40-60% less CPU usage on Apple Silicon since it's native WebKit.
https://github.com/achiya-automation/safari-mcp
80 tools for browser automation, MIT licensed. Works with Claude Code, Cursor, and Windsurf.
I have the same issue. Even when nothing is executing, CPU will be 75-100%
It is not fixed up to stable 2.1.153