Codex Desktop will not boot up on Mac OS Arm

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

• #19220

Powered by Codex Action

I have this same issue on mac.

Update: I was seeing the same issue earlier, but after signing out and signing back in, the app started working again.

I also have this same issue on mac

also have this same issue

also have this same issue

I resolved this problem by quitting Codex, removing ~/.codex/auth.json, then reopening Codex and logging in again.you can try it

I resolved this problem by quitting Codex, removing ~/.codex/auth.json, then reopening Codex and logging in again.you can try it

I have this same issue on mac.

I resolved this problem by quitting Codex, removing ~/.codex/auth.json, then reopening Codex and logging in again.you can try it

### What version of the Codex App are you using (From “About Codex” dialog)? Latest download as of 4/23/2026 ### What subscription do you have? Plus ### What platform is your computer? Darwin 24.6.0 arm64 arm ### What issue are you seeing? Following Error keeps popping up after boot up and after I press ok. `` Error: Connection for host ID local not found at Gd.getConnection (/Applications/Codex.app/Contents/Resources/app.asar/.vite/build/workspace-root-drop-handler-Cne_iOBj.js:247:13051) at db.getChronicleSidecarControlState (/Applications/Codex.app/Contents/Resources/app.asar/.vite/build/main-DPIxE_em.js:472:44880) at db.updateChronicleTrayIcon (/Applications/Codex.app/Contents/Resources/app.asar/.vite/build/main-DPIxE_em.js:420:21432) at Timeout._onTimeout (/Applications/Codex.app/Contents/Resources/app.asar/.vite/build/main-DPIxE_em.js:420:17842) at listOnTimeout (node:internal/timers:605:17) at process.processTimers (node:internal/timers:541:7) `` <img alt="Image" width="595" height="370" src="https://private-user-images.githubusercontent.com/148401030/583025384-c51103c6-1362-42d0-afa0-bcec11e9b06e.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3NzcwMDIzMzksIm5iZiI6MTc3NzAwMjAzOSwicGF0aCI6Ii8xNDg0MDEwMzAvNTgzMDI1Mzg0LWM1MTEwM2M2LTEzNjItNDJkMC1hZmEwLWJjZWMxMWU5YjA2ZS5wbmc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjYwNDI0JTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI2MDQyNFQwMzQwMzlaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT1iYjkzYzI2MDYzMzYwYzRhOTUxYmJmMWRkODhhZTYyMzliZDc5MDEyMWYxMjliMDllNjE1ODQwNzhiNmFlNmViJlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCZyZXNwb25zZS1jb250ZW50LXR5cGU9aW1hZ2UlMkZwbmcifQ.HkM3YjtNb_ZnwJyop_bca7XSdAiSZwfch7rcO_Jqq0o"> ### What steps can reproduce the bug? Install app. Start app. I uninstalled it and redid the steps above. ### What is the expected behavior? Starts up. ### Additional information _No response_

I resolved this problem by quitting Codex, removing ~/.codex/auth.json, then reopening Codex and logging in again.you can try it

I blew out the Brew Cask manually and reinstalled the CLI, then re-authed, and the issue was solved for me as well.

I blew out the Brew Cask manually and reinstalled the CLI, then re-authed, and the issue was solved for me as well.

I think it was an authentication issue.

I hit this too and asked the new 5.5 model to poke at the shipped desktop bundle a bit, this is what it came back with:

It looks like the immediate crash is not auth.json itself, but the tray / Chronicle polling path assuming the local app-server connection always exists:

updateChronicleTrayIcon()
-> getChronicleSidecarControlState()
-> appServerConnectionRegistry.getConnection("local")
-> throws "Connection for host ID local not found"

getConnection("local") is a throwing lookup.
[end codex]

My understanding is that if local startup/auth/config init fails, the macOS app blows up when it cannot reach the app server because the tray controller polls Chronicle state about once a second and throws each time. (I got multiple windows popping up with the exception and had to force quit the app).

The wiping auth.json fix likely works because it forces Codex down a different path to start the connection to the app server (i.e. with stale auth.json present Codex expects a working app server and throws without one but with no auth.json Codex enters a login/unauthenticated path that does not start the tray timer with getConnection("local") until authentication is achieved). At that point, Codex now has a valid auth.json so the app server connection succeeds.

I did a test in which I took a working auth.json and modified the refresh token to make it invalid and promptly hit the crash. codex logout / login resolved the issue.

Same issue here. Our entire org can't use Codex.

Same issue here. Our entire org can't use Codex.

@mzavattaro remove ~/.codex/auth.json and restart the app and it will ask you re-log in and will work

Open codex in terminal -> /logout -> open codex app again

I don't know whats happening but I solved this and claude let me to paste these text :

macOS, ChatGPT Plus, Codex desktop 26.422.30944. White screen with only the logo, Cmd+Opt+I did nothing, View menu had no Toggle Developer Tools, reinstall didn't help.

What worked for me: deleting the ~/.codex folder and re-logging in.

osascript -e 'tell application "Codex" to quit' 2>/dev/null; sleep 2; pkill -9 -f Codex 2>/dev/null; [ -d ~/.codex ] && mv ~/.codex ~/.codex.bak.$(date +%s) && echo "Renamed ~/.codex -> ~/.codex.bak.*"; open -a Codex

After this Codex launched normally, asked me to log in again, and the white screen was gone. Note ~/.codex is renamed (not deleted) so it's safe to revert if anything goes wrong.

Hope this helps someone else stuck on the same thing.

Hitting this exact crash on macOS ARM, Codex Desktop 26.422.62136 (also tried 26.422.30944). App hangs on splash; same getChronicleSidecarControlState → Connection for host ID local not found stack trace.

Plugin runtime auto-updated to 26.426.12240 today (openai-primary-runtime/{spreadsheets,documents,presentations}, openai-bundled/browser-use/0.1.0-alpha1) while the app is still on 26.422.x — looks like a plugin/app version skew.

Clearing Cache, GPUCache, Code Cache, Local Storage, Session Storage, plugin cache, and resetting .codex-global-state.json doesn't help. CLI works fine. Any ETA on a fix?

Confirming the auth.json fix worked for me — moved ~/.codex/auth.json aside, relaunched Codex, logged back in, app boots cleanly.

Heads-up for anyone hitting this: after re-login the chat list looked empty even though all data was intact on disk. Two extra steps were needed:

1. In ~/.codex/state_5.sqlite every thread had has_user_event=0. Fix:

``sql
UPDATE threads SET has_user_event=1 WHERE first_user_message != '';
``

1. The new .codex-global-state.json had no active-workspace-roots. UI scopes chats per workspace — File → Open Workspace and re-add the folders your threads were created in. Use this to find them:

``bash
sqlite3 ~/.codex/state_5.sqlite "SELECT cwd, COUNT(*) FROM threads GROUP BY cwd ORDER BY 2 DESC LIMIT 10;"
``

Thanks @Suchenovo @roborule for the original fix.

I have same issue on my macbook and unable to get this resolved. Stuck on a forever loop of authentication failed invalid_state.