Booting MCP Server
What version of Codex CLI is running?
codex-cli 0.120.0
What subscription do you have?
Plus with Codex Credit
Which model were you using?
gpt 5.4 Medium Fast
What platform is your computer?
Microsoft Windows NT 10.0.26200.0 x64
What terminal emulator and version are you using (if applicable)?
PowerShell
What issue are you seeing?
It says "Booting MCP server: codex_apps (8m 09s • esc to interrupt)" and I'm running out of credits, even though it's not doing anything; hundreds of credits are wasted.
<img width="908" height="143" alt="Image" src="https://github.com/user-attachments/assets/3a776e3d-5882-4bf8-975e-66619c22c1b3" />
What steps can reproduce the bug?
I don't know, but I added more than 10 queue messages in the CLI.
What is the expected behavior?
My loans are running out, but there's no response.
Additional information
It says "Booting MCP server: codex_apps (8m 09s • esc to interrupt)" and I'm running out of credits, even though it's not doing anything; hundreds of credits are wasted.
<img width="908" height="143" alt="Image" src="https://github.com/user-attachments/assets/3a776e3d-5882-4bf8-975e-66619c22c1b3" />
23 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
When you're waiting for a local MCP server to start, you're not sending anything to the model, so no tokens are used.
This is a bug, and how can you be so sure and speak so definitively without seeing it?
I was tracking my credits while writing MCP booting and I watched as about 200 credits were wasted.
If you're seeing credits being consumed, then there's more going on here than is described in your bug report above.
If you're able to repro, use
/feedbackto upload your logs and session details, then post the thread ID here.Uploaded thread: 019d6a4c-6734-7853-9a30-7f622a2b8ce4
Same issue.
Differences:
Same issue
But @etraut-openai says this is normal :)
same issue
thread id: 019dba89-73ca-7c72-a938-31d64be28cee
was running in macbook, vscode terminal
weird though. when i check /mcp
codex_apps seems like a dummy mcp with no tools.
why would i try to reboot this and get stuck on it?
my fix:
just prompted to continue the task without calling codex_apps mcp server. luckily, this worked and did not get into this bug loop.
Same issue. "This is normal" is not an explanation. Why does it do that?
This is exactly the problem why a lot of people not using Codex...
I am using Codex CLI and since 40 minutes it shows "Booting MCP Server"...
<img width="379" height="75" alt="Image" src="https://github.com/user-attachments/assets/98b31931-b1b0-4164-ac65-5aa26c8b1e1c" />
Kept happing to me, I got out of the tmux(psmux on windows) session and it was able to continue. I have not had problems with this until today.
Same for me:
• Booting MCP server: codex_apps (23m 30s • esc to interrupt)
Same for me: _Booting MCP server: codex_apps (8m 49s • esc to interrupt)_
Thread ID 019ded31-e27c-7223-9173-7f6b32138d60
why the fuck is this mcp being started if i aint using mcp at all
Are you all running in tmux session? I think this could be tmux specific.
Nope - running directly in GHOSTTY terminal window.
https://github.com/openai/codex/issues/17444
same for me
Yep, coming across the same issue. Seemingly at random it'll go into
Booting MCP server: codex_appsat which case the entire progress of the session just seems to halt... Don't have session on me right now so I can't leave a proper /feedback.linux mint 22.3
allacritty 0.17.0
codex-cli 0.129.0
tmux 3.4
No MCP servers installed, coincidentally removed my .codex dir a couple hours ago after the 0.129.0 update so I have pretty much a 100% fresh install, except for a couple config lines and 2 skills.
Edit: also happening on 0.130.0
Still no fix?
I found a temporary fix by adding this to my
~/.codex/config.toml: