You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
What terminal emulator and version are you using (if applicable)?
No response
Codex doctor report
codexError: host Job Object prevents daemon detachment; start from a host that allows breakawayTo work without the background server, rerun the same command with --no-daemon (including resume or fork and its arguments).
What issue are you seeing?
What steps can reproduce the bug?
just start
What is the expected behavior?
No response
Additional information
No response
Reacted by Diogo Mendonça, Đào Trung Thành, O Programador, Douglas Coimbra de Andrade, Jose M. Martinez, montague27, kevin-alt-acct, Skypix11, wupaz, Hiago Lopes and 13 more
Reacted by Barry Liang, BorzakA, Douglas Coimbra de Andrade, Kevin, Suraj Verma and joel-evercoast
I have encountered the same issue after the upgrade to 0.157.0.
PS ~\code> codex
Error: host Job Object prevents daemon detachment; start from a host that allows breakaway
To work without the background server, rerun the same command with --no-daemon (including resume or fork and its arguments).
PS ~\code> codex --version
codex-cli 0.157.0
PS ~\code>
Reacted by Hetav Pai Angle, Victor Zakharov and Nate-BadScienceFiction
PS path\to\project> codex
Error: host Job Object prevents daemon detachment; start from a host that allows breakaway
To work without the background server, rerun the same command with --no-daemon (including resume or fork and its arguments).
Getting the same here, codex 0.157.0 is installed under npm 11.19.0, Win 11 25H2 26200.8246
Error: host Job Object prevents daemon detachment; start from a host that allows breakaway
To work without the background server, rerun the same command with --no-daemon (including resume or fork and its arguments).
Windows Terminal 1.24 with Git Bash; the probes below also ran from PowerShell started outside Git Bash
Modern Standby supported (powercfg /a: "Standby (S0 Low Power Idle) Network Connected"), dam service running
Membership probe (first script below)
The caller's immediate job has LimitFlags=0x1800 (BREAKAWAY_OK and SILENT_BREAKAWAY_OK, no KILL_ON_JOB_CLOSE) and holds about 145 active processes. explorer.exe is in a job too, so the job is session-wide rather than added by the terminal.
A child created with CREATE_SUSPENDED | DETACHED_PROCESS | CREATE_BREAKAWAY_FROM_JOB is created fine, but IsProcessInJob(child, NULL) is TRUE. Same result with and without Git Bash in the chain.
Survival probe (second script below)
The child is spawned with the daemon's own flags from pid_start.rs (DETACHED_PROCESS | CREATE_BREAKAWAY_FROM_JOB) and checked from a separate observer process:
exit launcher exited=True (code 0); child 40636 at spawn: in-job; alive 5s after launcher gone: True; now: in-job
wait launcher exited=True (code -1073741510); child 8788 at spawn: in-job; alive 5s after launcher gone: True; now: in-job
exit: the launcher exits normally. The child keeps running.
wait: the launcher's console window gets WM_CLOSE (what the X button sends), and the launcher dies with STATUS_CONTROL_C_EXIT. The child keeps running.
So on this machine the residual job does not stop a detached child from outliving its launcher or its console. Not tested: behaviour across Modern Standby (DAM may throttle processes in its job), logoff, or closing a Windows Terminal tab (the probe used a hidden conhost window).
Suggestion: now that 0.157.0 starts the daemon by default, a failed detachment check makes Codex unusable instead of only disabling codex agents. Falling back to in-process mode with a warning (what --no-daemon does) would avoid that, however the job check itself ends up.
Workaround in ~/.codex/config.toml:
[features]
daemon_auto_start = false
Membership probe (PowerShell)
Add-Type-TypeDefinition @"using System;using System.Runtime.InteropServices;public static class B { [StructLayout(LayoutKind.Sequential)] public struct SI { public int cb; public IntPtr r, d, t; public int x, y, xs, ys, xc, yc, fa, fl; public short sw, cb2; public IntPtr r2, i, o, e; } [StructLayout(LayoutKind.Sequential)] public struct PI { public IntPtr hP, hT; public int pid, tid; } [StructLayout(LayoutKind.Sequential)] public struct BASIC { public long a, b; public uint LimitFlags; public UIntPtr c, d; public uint e; public UIntPtr f; public uint g, h; } [StructLayout(LayoutKind.Sequential)] public struct IOC { public ulong a, b, c, d, e, f; } [StructLayout(LayoutKind.Sequential)] public struct EXT { public BASIC Basic; public IOC Io; public UIntPtr p, j, pp, jp; } [StructLayout(LayoutKind.Sequential)] public struct ACCT { public long a, b, c, d; public uint TotalPageFaultCount, TotalProcesses, ActiveProcesses, TotalTerminatedProcesses; } [DllImport("kernel32.dll", SetLastError=true, CharSet=CharSet.Unicode)] public static extern bool CreateProcessW(string app, System.Text.StringBuilder cmd, IntPtr pa, IntPtr ta, bool inh, uint flags, IntPtr env, string cwd, ref SI si, out PI pi); [DllImport("kernel32.dll", SetLastError=true)] public static extern bool IsProcessInJob(IntPtr p, IntPtr j, out bool r); [DllImport("kernel32.dll", SetLastError=true)] public static extern bool QueryInformationJobObject(IntPtr j, int cls, out EXT info, int len, IntPtr ret); [DllImport("kernel32.dll", SetLastError=true)] public static extern bool QueryInformationJobObject(IntPtr j, int cls, out ACCT info, int len, IntPtr ret); [DllImport("kernel32.dll")] public static extern bool TerminateProcess(IntPtr h, uint c); [DllImport("kernel32.dll")] public static extern bool CloseHandle(IntPtr h); public static string Run() { string o = ""; EXT e; if (QueryInformationJobObject(IntPtr.Zero, 9, out e, Marshal.SizeOf(typeof(EXT)), IntPtr.Zero)) { uint f = e.Basic.LimitFlags; o += string.Format("own job LimitFlags=0x{0:X} BREAKAWAY_OK={1} SILENT_BREAKAWAY_OK={2} KILL_ON_CLOSE={3}\n", f, (f & 0x800) != 0, (f & 0x1000) != 0, (f & 0x2000) != 0); } else o += "own job query failed err=" + Marshal.GetLastWin32Error() + "\n"; ACCT a; if (QueryInformationJobObject(IntPtr.Zero, 1, out a, Marshal.SizeOf(typeof(ACCT)), IntPtr.Zero)) o += string.Format("own job ActiveProcesses={0} TotalProcesses={1}\n", a.ActiveProcesses, a.TotalProcesses); SI si = new SI(); si.cb = Marshal.SizeOf(typeof(SI)); PI pi; // same flags as codex ensure_detached_launch: CREATE_SUSPENDED | DETACHED_PROCESS | CREATE_BREAKAWAY_FROM_JOB uint flags = 0x4 | 0x8 | 0x01000000; if (!CreateProcessW(@"C:\Windows\System32\cmd.exe", new System.Text.StringBuilder("cmd.exe /c exit"), IntPtr.Zero, IntPtr.Zero, false, flags, IntPtr.Zero, null, ref si, out pi)) return o + "breakaway CreateProcess FAILED err=" + Marshal.GetLastWin32Error(); bool inJob; IsProcessInJob(pi.hP, IntPtr.Zero, out inJob); TerminateProcess(pi.hP, 0); CloseHandle(pi.hT); CloseHandle(pi.hP); return o + "breakaway child in job after CREATE_BREAKAWAY_FROM_JOB: " + inJob; }}"@
[B]::Run()
Survival probe (PowerShell)
# Survival probe for openai/codex#44503. Launcher mode spawns a long-lived child with the# daemon's own flags (DETACHED_PROCESS | CREATE_BREAKAWAY_FROM_JOB), records it, then# either exits or waits for its console window to be closed. Observer mode drives both.param([string]$Mode='observe', [string]$Out="$env:TEMP\codex-lifecycle")
Add-Type-TypeDefinition @"using System;using System.Runtime.InteropServices;public static class L { [StructLayout(LayoutKind.Sequential)] public struct SI { public int cb; public IntPtr r, d, t; public int x, y, xs, ys, xc, yc, fa, fl; public short sw, cb2; public IntPtr r2, i, o, e; } [StructLayout(LayoutKind.Sequential)] public struct PI { public IntPtr hP, hT; public int pid, tid; } [DllImport("kernel32.dll", SetLastError=true, CharSet=CharSet.Unicode)] public static extern bool CreateProcessW(string app, System.Text.StringBuilder cmd, IntPtr pa, IntPtr ta, bool inh, uint flags, IntPtr env, string cwd, ref SI si, out PI pi); [DllImport("kernel32.dll", SetLastError=true)] public static extern IntPtr OpenProcess(uint a, bool i, int pid); [DllImport("kernel32.dll", SetLastError=true)] public static extern bool IsProcessInJob(IntPtr p, IntPtr j, out bool r); [DllImport("kernel32.dll")] public static extern bool CloseHandle(IntPtr h); [DllImport("kernel32.dll")] public static extern IntPtr GetConsoleWindow(); [DllImport("user32.dll")] public static extern bool PostMessageW(IntPtr h, uint m, IntPtr w, IntPtr l); public static int Spawn() { SI si = new SI(); si.cb = Marshal.SizeOf(typeof(SI)); PI pi; // DETACHED_PROCESS | CREATE_BREAKAWAY_FROM_JOB, as in app-server-daemon/src/backend/pid_start.rs if (!CreateProcessW(@"C:\Windows\System32\PING.EXE", new System.Text.StringBuilder("PING.EXE -n 120 127.0.0.1"), IntPtr.Zero, IntPtr.Zero, false, 0x8 | 0x01000000, IntPtr.Zero, null, ref si, out pi)) throw new Exception("CreateProcess failed " + Marshal.GetLastWin32Error()); CloseHandle(pi.hT); CloseHandle(pi.hP); return pi.pid; } public static string InJob(int pid) { IntPtr h = OpenProcess(0x1000, false, pid); if (h == IntPtr.Zero) return "gone"; bool r; IsProcessInJob(h, IntPtr.Zero, out r); CloseHandle(h); return r ? "in-job" : "no-job"; }}"@if ($Mode-ne'observe') {
$child= [L]::Spawn()
"$child$([L]::InJob($child))$([L]::GetConsoleWindow().ToInt64())"|Set-Content"$Out-$Mode.txt"if ($Mode-eq'wait') { while ($true) { Start-Sleep1 } }
exit0
}
$self=$MyInvocation.MyCommand.Pathforeach ($min'exit','wait') {
Remove-Item"$Out-$m.txt"-ErrorAction SilentlyContinue
$launcher=Start-Processpowershell.exe-WindowStyle Hidden-PassThru `-ArgumentList '-NoProfile','-ExecutionPolicy','Bypass','-File',"`"$self`"",'-Mode',$mwhile (-not (Test-Path"$Out-$m.txt")) { Start-Sleep-Milliseconds 200 }
Start-Sleep-Milliseconds 500$childPid,$jobAtSpawn,$hwnd= (Get-Content"$Out-$m.txt").Split('')
if ($m-eq'wait') {
# Close the launcher's console window the way the X button does (WM_CLOSE = 0x10).
[void][L]::PostMessageW([IntPtr][long]$hwnd,0x10, [IntPtr]::Zero, [IntPtr]::Zero)
}
$launcher.WaitForExit(15000) |Out-NullStart-Sleep-Seconds 5$alive= [bool](Get-Process-Id $childPid-ErrorAction SilentlyContinue)
"{0,-5} launcher exited={1} (code {2}); child {3} at spawn: {4}; alive 5s after launcher gone: {5}; now: {6}"-f`$m,$launcher.HasExited,$launcher.ExitCode,$childPid,$jobAtSpawn,$alive, [L]::InJob([int]$childPid)
Stop-Process-Id $childPid-Force -ErrorAction SilentlyContinue
}
Reacted by moonrailgun, Joe Clinton, Cameron McCormick, Jared Miller, Victor Zakharov and Luís Delgado
I tried this and codex opens but at the first prompt I get
- The first shell launch failed before running the inspection commands, so I’m retrying with a simpler repository read.
- I couldn’t inspect or update _sut: the shell failed to start on each attempt, so I couldn’t read the test or service constructor. No files were changed. Please retry once the workspace shell is available
I can confirm the same issue after upgrading to Codex CLI 0.157.0 on Windows.
Normal startup fails:
codex
→ fails during startup
Starting with the daemon disabled works normally:
codex --no-daemon
→ starts successfully
So, at least in my environment, the failure appears isolated to the new background-server/daemon startup path introduced/enabled in 0.157.0. The interactive Codex session itself works when the daemon is bypassed.
I'm experiencing exactly the same issue with Codex CLI 0.157.0 on Windows, using PowerShell.
Steps to reproduce:
Open PowerShell.
Run codex.
Actual behavior:
Error: host Job Object prevents daemon detachment; start from a host that allows breakaway
To work without the background server, rerun the same command with --no-daemon (including resume or fork and its arguments).
Expected behavior:
Codex should start normally, as before, or automatically fall back to non-daemon mode if daemon detachment is unavailable.
I can reproduce this on Windows x64 with Codex CLI 0.157.0, and I did some additional Windows Job Object diagnostics that may help narrow the regression.
Environment
Hardware: Microsoft Surface Laptop 3
OS: Windows 11 Pro x64
Version/build: 10.0.26200 / build 26200.9457
Power model: Modern Standby — Standby (S0 Low Power Idle), Network Connected
HypervisorPresent: True
Virtualization-based security: Running
Codex installation: official standalone Windows PowerShell installer
Reproduction on 0.157.0
> codex --version
codex-cli 0.157.0
> codex
Error: host Job Object prevents daemon detachment; start from a host that allows breakaway
To work without the background server, rerun the same command with --no-daemon (including resume or fork and its arguments).
> codex app-server daemon start
Error: host Job Object prevents daemon detachment; start from a host that allows breakaway
The same failure occurs from both PowerShell 7.6.6 and plain cmd.exe.
codex --no-daemon works.
I also performed a clean reinstall of 0.157.0 using the standalone installer; the behavior was unchanged.
Windows Job Object diagnostics
Using IsProcessInJob confirms the shell is in a Job Object:
ApiCallSucceeded : True
ProcessIsInJob : True
However, querying JobObjectExtendedLimitInformation reports:
So the host Job Object has JOB_OBJECT_LIMIT_BREAKAWAY_OK set. Despite that, 0.157.0 reports that the host prevents daemon detachment.
Regression test
I downgraded the same machine to:
codex-cli 0.155.0
Plain codex then starts normally without --no-daemon.
No Windows/VBS/Hyper-V/Modern Standby configuration was changed between the 0.157.0 failure and the successful 0.155.0 test.
This looks like a regression between 0.155.0 and 0.157.0 in the Windows daemon detachment/Job Object handling. In particular, it may be worth checking whether the newer daemon startup path is treating presence in a Job Object as non-breakaway even when JOB_OBJECT_LIMIT_BREAKAWAY_OK is present.
Weird enough I see terminals spawning whatever command codex is running, does that happen to anyone?
I experience this too, running 0.157.1 in pwsh 7.6.6 on Windows 11. It makes it very hard to use codex CLI. codex --no-daemon doesn't have this behavior though.
Fwiw this worked to downgrade it for me
$env:CODEX_RELEASE = '0.156.1'; powershell -ExecutionPolicy ByPass -c "irm
https://chatgpt.com/codex/install.ps1 | iex"
PS C:\Users\Admin\Documents\Project> codex
Installing daemon from CLI version 0.157.1 into C:\Users\Admin\.codex\packages/app-server-daemon...
Error: Access is denied. (os error 5)
To work without the background server, rerun the same command with --no-daemon (including resume or fork and its arguments).
i can confirm this is still the case on 157.1 and is fixed by downgrading to 0.156, im running on windows 10
> codex --version
codex-cli 0.157.1
> codex
Error: cannot launch detached daemon; existing daemon was not stopped: The requested operation requires elevation. (os error 740)
To work without the background server, rerun the same command with --no-daemon (including resume or fork and its arguments).
running the cmd prompt/powershell as admin does not fix it and neither does setting this checkbox
in the codex binaries folder
please reopen this issue as i believe it is affecting most windows users
@etraut-openai
I just upgraded to 0.158.0 and got this warning
Then I try to upgrade the daemon and end up with this issue
>codex
Updating the local background server...
Error: host Job Object prevents daemon detachment; start from a host that allows breakaway
Error: Daemon update failed with status exit code: 1
Activity
24online24 commentedon Sep 25, 2026
I have encountered the same issue after the upgrade to 0.157.0.
vidi42 commentedon Sep 25, 2026
Just upgraded to 0.157.0 and got the same
sjasct commentedon Sep 25, 2026
Getting the same here, codex 0.157.0 is installed under npm 11.19.0, Win 11 25H2 26200.8246
model-bender commentedon Sep 25, 2026
Same here after updating to 0.157.0: plain
codexexits with this error. Adding data that answers the survival question asked in #44503.Environment
@openai/codex0.157.0,aarch64-pc-windows-msvcbinary)powercfg /a: "Standby (S0 Low Power Idle) Network Connected"),damservice runningMembership probe (first script below)
LimitFlags=0x1800(BREAKAWAY_OK and SILENT_BREAKAWAY_OK, no KILL_ON_JOB_CLOSE) and holds about 145 active processes.explorer.exeis in a job too, so the job is session-wide rather than added by the terminal.CREATE_SUSPENDED | DETACHED_PROCESS | CREATE_BREAKAWAY_FROM_JOBis created fine, butIsProcessInJob(child, NULL)is TRUE. Same result with and without Git Bash in the chain.Survival probe (second script below)
The child is spawned with the daemon's own flags from
pid_start.rs(DETACHED_PROCESS | CREATE_BREAKAWAY_FROM_JOB) and checked from a separate observer process:exit: the launcher exits normally. The child keeps running.wait: the launcher's console window getsWM_CLOSE(what the X button sends), and the launcher dies with STATUS_CONTROL_C_EXIT. The child keeps running.So on this machine the residual job does not stop a detached child from outliving its launcher or its console. Not tested: behaviour across Modern Standby (DAM may throttle processes in its job), logoff, or closing a Windows Terminal tab (the probe used a hidden conhost window).
Suggestion: now that 0.157.0 starts the daemon by default, a failed detachment check makes Codex unusable instead of only disabling
codex agents. Falling back to in-process mode with a warning (what--no-daemondoes) would avoid that, however the job check itself ends up.Workaround in
~/.codex/config.toml:Membership probe (PowerShell)
Survival probe (PowerShell)
moonrailgun commentedon Sep 25, 2026
same here and
work for me
gitPhate commentedon Sep 25, 2026
I tried this and codex opens but at the first prompt I get
nikolaidk commentedon Sep 25, 2026
I have similar issue on MAC
Transient-Onlooker commentedon Sep 25, 2026
tleneel commentedon Sep 25, 2026
I just updated the codex as it requested an updated. And upon return I have the same error :/
Here is the fix : npx --yes @openai/codex@0.156.0 → DO NOT accept to upgrade to 0.157, press "3 - skip this version" :)
Bolt81 commentedon Sep 25, 2026
I can confirm the same issue after upgrading to Codex CLI 0.157.0 on Windows.
Normal startup fails:
Starting with the daemon disabled works normally:
So, at least in my environment, the failure appears isolated to the new background-server/daemon startup path introduced/enabled in 0.157.0. The interactive Codex session itself works when the daemon is bypassed.
Environment:
codex --no-daemonthanh-abaii commentedon Sep 25, 2026
I'm experiencing exactly the same issue with Codex CLI 0.157.0 on Windows, using PowerShell.
Steps to reproduce:
Open PowerShell.
Run codex.
Actual behavior:
Error: host Job Object prevents daemon detachment; start from a host that allows breakaway
To work without the background server, rerun the same command with --no-daemon (including resume or fork and its arguments).
Expected behavior:
Codex should start normally, as before, or automatically fall back to non-daemon mode if daemon detachment is unavailable.
This issue appeared after upgrading to 0.157.0.
10 remaining items
Utahleo commentedon Sep 25, 2026
I can reproduce this on Windows x64 with Codex CLI 0.157.0, and I did some additional Windows Job Object diagnostics that may help narrow the regression.
Environment
Reproduction on 0.157.0
The same failure occurs from both PowerShell 7.6.6 and plain cmd.exe.
codex --no-daemonworks.I also performed a clean reinstall of 0.157.0 using the standalone installer; the behavior was unchanged.
Windows Job Object diagnostics
Using
IsProcessInJobconfirms the shell is in a Job Object:However, querying
JobObjectExtendedLimitInformationreports:So the host Job Object has
JOB_OBJECT_LIMIT_BREAKAWAY_OKset. Despite that, 0.157.0 reports that the host prevents daemon detachment.Regression test
I downgraded the same machine to:
Plain
codexthen starts normally without--no-daemon.No Windows/VBS/Hyper-V/Modern Standby configuration was changed between the 0.157.0 failure and the successful 0.155.0 test.
This looks like a regression between 0.155.0 and 0.157.0 in the Windows daemon detachment/Job Object handling. In particular, it may be worth checking whether the newer daemon startup path is treating presence in a Job Object as non-breakaway even when
JOB_OBJECT_LIMIT_BREAKAWAY_OKis present.jzhu230 commentedon Sep 26, 2026
I have encountered the same issue, and I solved it by telling codex (using the GUI) to downgrade to 0.156.1
borzaka commentedon Sep 26, 2026
Looks like it's fixed in 0.157.1, the CLI now starts.
gitPhate commentedon Sep 26, 2026
Weird enough I see terminals spawning whatever command codex is running, does that happen to anyone?
abarefoot commentedon Sep 26, 2026
I experience this too, running 0.157.1 in pwsh 7.6.6 on Windows 11. It makes it very hard to use codex CLI.
codex --no-daemondoesn't have this behavior though.kevin-alt-acct commentedon Sep 26, 2026
etraut-openai commentedon Sep 26, 2026
We published 0.157.1 as a hot-fix release to address this issue. If you're still seeing this problem after updating, let us know.
tleneel commentedon Sep 26, 2026
loading, but with issues. It spam by itself text like :
[M#x1[M#u1[M#q1[M#m1[M#l1[M#i1[M#h1[M#d1[M#c1[M#
1[M#_1[M#]1[M#\1[M#X1[M#U1[M#S2[M#R2[M#Q2[M#P2[M#N2[M#M2[M#L2[M#K2[M#J2[M#J3[M#O2[M#O1[M#N1[M#N0[M#N/[M#O/[M#P/ [M#Q/[M#R/[M#Q/[M#P/[M#M.[M#J.[M#G.[M#E.[M#D.[M#C.[M#B.[M#A.[M#?.[M#>.[M#;.[M#9.[M#8.[M#7.[M#6.[M#5.[M#4.[M#3.[M#2.[M#1.[M#0.[M#/.[M#..[M#-.[M#,.[M#+.[M#*.[M#*/[M#+/ [M#,0[M#-0[M#.0[M#/0[M#//[M#0/[M#1/[M#1.[M#0.[M#/.[M#..[M 3,[M#3,[M#3,[M#3-[M 3-[M@5-[M@7-[M@9-[M@:-[M@<,[M@@,[M@C,[M@F,[M@I,[M@L,[M@M,[M@S,[M@V,[M@X,[M@[,[M@\,[M@],[M@_,[M@,[M@a,[M@c,[M@d,[M@f,[M@g-[M@h-[M@j-[M@k-[M@l-[M@l.[M@m.[M@n/[M@o/[M@p/[M@q/[M@r/[M@s/[M@t/[M@t0[M@u0[M@v0[M@w0[M@x0[M@y0[M@z0[M@{0[M@|0[M@}0[M#}0[M#|
0[M#{0[M#z0[M#y0[M#v0[M#t0[M#s0[M#p0[M#o0[M#n0[M#m0[M#m/[M m/[M#Z,[M#X,[M#V,[M#U,[M#T,[M#S,[M#R,[M#Q,[M#P,[M#O,[M#N,[M#M,[M#L,[M#K,[M#J,[M#I,[M#H,[M#G,[M#F,
[M#F+[M#E+[M#D+[M#C+[M#B+[M#A+[M#@+[M#?+[M#>+[M#=+[M#<+[M#;+[M#:+[M#9+[M#8+[M#8,[M#6,[M#4,[M#3,[M#2,[M#1,[M#0,[M#0+[M#/+[M#.+[M#-+[M#,+[M#.+[M#/+[M#0+[M#1+[M#1*[M#2*
etraut-openai commentedon Sep 26, 2026
@tleneel, please file a new bug report with the details requested in the bug template.
gravitydeepimpactsun commentedon Sep 27, 2026
@etraut-openai Still happening after updating.
lain804 commentedon Sep 27, 2026
@etraut-openai
i can confirm this is still the case on 157.1 and is fixed by downgrading to 0.156, im running on windows 10
running the cmd prompt/powershell as admin does not fix it and neither does setting this checkbox
in the codex binaries folder
please reopen this issue as i believe it is affecting most windows users
radmi03 commentedon Sep 28, 2026
Confirming the Windows regression is still reproducible on Codex CLI 0.157.1.
Observed on Windows:
codex->Installing daemon from CLI version 0.157.1 ...thenError: Access is denied. (os error 5)codex --no-daemonworksCODEX_HOME\packagessucceedapp-server-daemon\releasesandinstall.lock, but no daemon binary is extracted before the failure@openai/codex@0.156.1fixes the issue; normalcodexwith daemon works againThis strongly suggests a 0.157.x Windows daemon installation/startup regression rather than a local NTFS permission problem.
vidi42 commentedon Sep 28, 2026
@etraut-openai
I just upgraded to 0.158.0 and got this warning
Then I try to upgrade the daemon and end up with this issue
leobrqz commentedon Sep 28, 2026
same issue in the latest update