Codex fails completely in WSL

Resolved 💬 2 comments Opened Jan 20, 2026 by dirkknibbe Closed Feb 5, 2026
💡 Likely answer: A maintainer (etraut-openai, contributor) responded on this thread — see the highlighted reply below.

What version of Codex is running?

codex-cli 0.87.0

What subscription do you have?

Enterprise

Which model were you using?

gpt-5.2-codex high

What platform is your computer?

Microsoft Windows NT 10.0.26100.0 x64

What terminal emulator and version are you using (if applicable)?

Windows Terminal with Ubuntu 22.04.5 LTS

What issue are you seeing?

no matter the prompt it always hangs forever and then loses connection and tries to reconnect

What steps can reproduce the bug?

Uploaded thread: 019bdc13-2b60-79c2-8eb5-69000eb2f220

What is the expected behavior?

I would expect to get some output

Additional information

_No response_

View original on GitHub ↗

2 Comments

etraut-openai contributor · 7 months ago

@dirkknibbe, your logs were not uploaded, so I don't have much information to help you. If you're able to repro this problem, please try to use the /feedback slash command again and post the thread ID here.

You mentioned that it loses connection and tries to reconnect. Can you provide the error message that you see?

Was this working in a previous version of Codex, or is this the first time you're using it?

Are you logging in with an API Key or a Chat GPT subscription? The behavior you're describing could be explained if you logged in with an invalid API key. Try logging out and logging back in and confirm that the key is correct.

Are you using an OpenAI (default) endpoint, or are you using an Azure-hosted model?

Do you have any other non-default settings in your config?

Do you have a firewall, VPN, proxy or other special network configurations on your machine or enterprise network?

dirkknibbe · 6 months ago

I'm sorry yea I discovered this is a issue with our enterprise Cisco Umbrella DNS rules resolving WSL codex traffic to blocked IP's whereas windows is resolving properly, closing this issue