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
I am trying to understand the intended support boundary for Codex Desktop when using a WSL-side agent from the Windows Desktop app.
I searched existing issues and Discussions before posting. I found several symptom-specific issues, but I did not find an umbrella Discussion about the broader pattern.
The pattern I am seeing is:
Multiple Codex Desktop features that require Windows-to-WSL coordination appear to fail or degrade in WSL agent mode.
This is not meant as a duplicate bug report, and I am not claiming every WSL Desktop user is affected. The concern is that several separate open issues seem to point at the same class of boundary problem: the Windows Desktop shell, app-server, helper runtimes, browser tooling, MCP launch environment, Git state, and WSL filesystem/config state do not always agree about where execution is happening.
The MCP case is particularly useful as a lower-level example. In one current WSL Desktop setup, a Playwright MCP server was configured in the WSL Codex home, but the Desktop-launched environment resolved npx through Windows paths rather than the intended WSL Node/NVM path. Current source suggests that once an MCP server starts and lists tools successfully, those tools should be exposed directly or via tool search. So missing tools in that setup look more like a launch/readiness/environment problem than a user simply looking in the wrong place.
My question is:
Is WSL agent mode intended to support Desktop features like automations, Browser Use / @Chrome, Git/PR status, workspace dependencies, and custom MCP servers reliably today?
If yes, is there a recommended configuration for Windows Desktop + WSL that avoids these boundary failures?
If no, could the current support boundary be documented more explicitly, so users know when to prefer the WSL CLI over Desktop?
The reason I am asking this as a Discussion rather than opening another issue is that the individual bug reports already exist. The collective impact is the part that seems easy to miss: for Windows users coding primarily inside WSL, these are some of the features that make Desktop attractive over the CLI, so failures across them materially affect the Desktop experience.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
-
I am trying to understand the intended support boundary for Codex Desktop when using a WSL-side agent from the Windows Desktop app.
I searched existing issues and Discussions before posting. I found several symptom-specific issues, but I did not find an umbrella Discussion about the broader pattern.
The pattern I am seeing is:
This is not meant as a duplicate bug report, and I am not claiming every WSL Desktop user is affected. The concern is that several separate open issues seem to point at the same class of boundary problem: the Windows Desktop shell, app-server, helper runtimes, browser tooling, MCP launch environment, Git state, and WSL filesystem/config state do not always agree about where execution is happening.
A few examples:
CODEX_HOME..gitpaths.@Chromeintegration issues on Windows Desktop, including missing Node REPL tool exposure and fallback to an isolated browser path.workspace_dependenciesfeature/runtime mismatch after local recovery.The MCP case is particularly useful as a lower-level example. In one current WSL Desktop setup, a Playwright MCP server was configured in the WSL Codex home, but the Desktop-launched environment resolved
npxthrough Windows paths rather than the intended WSL Node/NVM path. Current source suggests that once an MCP server starts and lists tools successfully, those tools should be exposed directly or via tool search. So missing tools in that setup look more like a launch/readiness/environment problem than a user simply looking in the wrong place.My question is:
Is WSL agent mode intended to support Desktop features like automations, Browser Use /
@Chrome, Git/PR status, workspace dependencies, and custom MCP servers reliably today?If yes, is there a recommended configuration for Windows Desktop + WSL that avoids these boundary failures?
If no, could the current support boundary be documented more explicitly, so users know when to prefer the WSL CLI over Desktop?
The reason I am asking this as a Discussion rather than opening another issue is that the individual bug reports already exist. The collective impact is the part that seems easy to miss: for Windows users coding primarily inside WSL, these are some of the features that make Desktop attractive over the CLI, so failures across them materially affect the Desktop experience.
Beta Was this translation helpful? Give feedback.
All reactions