⚠️ stream error: stream disconnected before completion: Transport error: error decoding response body; retrying 1/5 in 199ms…
Resolved 💬 12 comments Opened Sep 16, 2025 by 3SMMZRjWgS Closed Oct 16, 2025
What version of Codex is running?
codex-cli 0.36.0
Which model were you using?
gpt-5-codex high
What platform is your computer?
Windows 11 Pro
What steps can reproduce the bug?
For the latest version of the Codex CLI, if you run it with VPN on, you'll get the stream disconnected error pretty much in the first couple of chats.
What is the expected behavior?
Connection should be agnostic of user's choice of using a VPN or not as it has been with Codex CLI up to 0.34.0
What do you see instead?
I see this whenever I run a Codex CLI session with my NordVPN on:⚠️ stream error: stream disconnected before completion: Transport error: error decoding response
body; retrying 1/5 in 199ms…"
Additional information
_No response_
12 Comments
Same issue here, my organisation is verified, so it shouldn't be that
Same problem here. I'm on the API and have no VPN on.
Hitting a similar _stream error_, but seems to be caused by Your organization must be verified to use the model
gpt-5-codex. With or without VPN.We use the OpenAI account setup/managed by our parent company, so this is not something we can easily get changed 🤷🏼.
gpt-5has no such issues.Same error here. As described by OP. Running with and without VPN.
Same here
<img width="1137" height="295" alt="Image" src="https://github.com/user-attachments/assets/d8d8bc40-5f12-407f-a226-b6a114d6740e" />
same here
Same problem here.
Use gpt-5-codex-high will have a high chance to get this problem
I solved it! It's a token issue. Just run codex login and it works
I had a related problem to what's described in the issue when running Codex inside a Ubuntu VM with network mode NAT on a Windows host. Some requests would succeed, while others would not. Longer or more difficult prompts had the tendency to fail more:
Setting the VM's network adapter to bridged mode solved the issue for me. No more transport errors.
I guess I'll have to wait
"For developers using Codex CLI via API key, we plan to make GPT‑5-Codex available in the API soon." from: https://openai.com/index/introducing-upgrades-to-codex/
This bug is back in 0.42.0 on MacOs. Seemed fixed in the past few versions.
给中国用户的一个提示,如果显示这个错误,并且cli中也运行不了,那么应该是token过期或者mcp超时引起的,把.codex/config.toml里面的mcp配置删除,随后删除.codex/auth.json,运行codex重新登录,就能成功了!