App exhibits high memory usage
Open 💬 5 comments Opened Mar 14, 2026 by KnowhereFern
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
What version of the Codex App are you using (From “About Codex” dialog)?
Version 26.311.21342 (993)
What subscription do you have?
Pro $20
What platform is your computer?
Mac OS X 25.3.0 arm64 arm
What issue are you seeing?
Codex is behaving like a crackhead and is leaking ram. This is the second time in happens in 48 hours.
<img width="392" height="458" alt="Image" src="https://github.com/user-attachments/assets/8fc1e7d5-cdf5-40ed-8b61-8168b95314d9" />
What steps can reproduce the bug?
Use it normally, nothing special. I asked it to create GPT 6.0 make no mistakes and it happened like the attached,
What is the expected behavior?
For it to exist without consuming all my system ram. Ram is for millionaires these days smh.
Additional information
_No response_
5 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
I’m experiencing this too, multiple times a day until my Mac freezes and automatically reboots. It started after the last or second-to-last update. I didn’t have this issue before, and I’ve been using the app since launch (and really enjoying it!). Not doing anything crazy, no MCP's, just regular dev work. I get the feeling it happens around
bashtool calls, in my case, running Swift tests in bash.Same. Does codex provide any way to diagnose these things? Could be an MCP server, or a codex-internal thing, that is leaking memory.
On my side, generally the symptoms are:
I'm on an M1 Mac with the latest version of macOS.
I've been getting the oom dialog like the op posted as well. I run Codex hard, with multiple concurrent agents and MCP servers, etc., so I'm not _complaining_ exactly, but it would be _nice_ if the out of mem errors stopped happening.
I have been experiencing the same. I have two MCP servers that I use. Context7 and expect.dev . I do have a few skills/workflows that spawn multiple sub agents (plan review, code review, etc). What I am noticing is that when those subagents die, they are not killing the mcp servers initializations they did. I have to manually go and run the following command, then RAM usage goes down to ~1.5 GB vs running on 20+ GB of RAM before.
This happens to me on the mac os codex app. Running version 26.506.31421 and codex-cli 0.130.0