The reconnecting issue is even worse than last days.

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

• #14062
• #13519
• #13832
• #12458

Powered by Codex Action

I agree this is unusable.

Same issue today.
Version 26.305.950 (863)
MacOS
Model: GPT-5.4 Extra High

Reconnecting... 1/5
Reconnecting... 2/5
Explored 1 file
Reconnecting... 1/5
Reconnecting... 2/5
Reconnecting... 3/5
Reconnecting... 4/5

Same here. Here's a screenshot of what I'm seeing. I am on NordVPN, but it doesn't seem to make a difference if I turn it off.

App version: Version 26.305.950 (863)

<img width="1060" height="730" alt="Image" src="https://github.com/user-attachments/assets/a5fd5e91-53c3-4abf-ae6a-49ed312324cc" />

Most of these that we've investigated have turned out to be network connectivity issues (flaky networks, VPNs, proxies, firewalls, etc.).

We're looking at implementing a diagnostic utility to help users self-diagnose their networking problems.

I may be able to help you diagnose if you use /feedback to upload your logs and then post your thread ID here.

Most of these that we've investigated have turned out to be network connectivity issues (flaky networks, VPNs, proxies, firewalls, etc.). We're looking at implementing a diagnostic utility to help users self-diagnose their networking problems. I may be able to help you diagnose if you use /feedback to upload your logs and then post your thread ID here.

Hi @etraut-openai, with all due respect, are you sure? All of the sudden everyone is having "network connectivity issues (flaky networks, VPNs, proxies, firewalls, etc.)"?

Most of these that we've investigated have turned out to be network connectivity issues (flaky networks, VPNs, proxies, firewalls, etc.). We're looking at implementing a diagnostic utility to help users self-diagnose their networking problems. I may be able to help you diagnose if you use /feedback to upload your logs and then post your thread ID here.

We really appreciate your help but I do not agree with this: everything started 2-3 days ago exactly with the increase/abnormal usage.

Just saw https://status.openai.com/ and it seems like there is some known issue for Codex

<img width="723" height="696" alt="Image" src="https://github.com/user-attachments/assets/6add7f77-1418-4dba-a928-b5818fa1fd87" />

> Most of these that we've investigated have turned out to be network connectivity issues (flaky networks, VPNs, proxies, firewalls, etc.). > We're looking at implementing a diagnostic utility to help users self-diagnose their networking problems. > I may be able to help you diagnose if you use /feedback to upload your logs and then post your thread ID here. Hi @etraut-openai, with all due respect, are you sure? All of the sudden everyone is having "network connectivity issues (flaky networks, VPNs, proxies, firewalls, etc.)"?

Hi @etraut-openai , need to echo this sentiment, i am also using CC and Gemini cli, no reconnection issues there, whereas with the Codex CLI I have probably gotten 100+ the last 4 hours, completely un-useable at the moment. On the Pro plan.

Please use /feedback and upload your logs, then post your thread ID here. Without logs, there's very little I can do to help.

@etraut-openai 019cd812-ca4e-76e0-b6c6-6fc3eb7bb96b

Same problem. Few days ago everything was perfect, even with vpns/corporate proxies etc.
For now its just unusable, first Falling back from WebSockets to HTTPS transport then reconnecting every few steps.

Rollback to older version fixes it like a magic.

@nprange, I don't see any reconnection attempts in your logs. Did you repro the issue prior to using /feedback?

@omggga, I don't think that's the same problem being reported here. It sounds like you are on a network that is blocking websockets for some reason. Perhaps a corporate firewall, VPN or proxy?

@nprange, I don't see any reconnection attempts in your logs. Did you repro the issue prior to using /feedback?

Yes I did, how strange, it should've appeared for you @etraut-openai, but here is a printscreen of it:

<img width="417" height="289" alt="Image" src="https://github.com/user-attachments/assets/5c62f5d9-7578-458f-a5ce-2134c960bf16" />

<img width="759" height="86" alt="Image" src="https://github.com/user-attachments/assets/f53cc452-c158-44b9-a449-41edaaed0eea" />

Personal Mac Mini, no proxy, no VPN.

I’m having a persistent issue with Codex Desktop on Windows.

first 2 conversations are stuck in an endless loading/spinning state. then all conversations having same problem. Restarting my PC and reopening Codex does not fix it. As soon as the app opens, the same conversations start spinning again, but nothing actually runs and Stop does not work.

From the local logs, the conversations are repeatedly resumed as if they are still in progress, but backend actions fail with "no active turn to steer". I also see app-server/auth-related errors, including "invalid_token", which suggests the local session/thread state became inconsistent.

Affected conversation IDs:

• 019c9a9b-0faa-74e2-ad1d-07bdd62a9d56
• 019c8e1a-3737-77d0-bff0-cf8a652c1b14

@cociboni, that is not the same problem being reported in this issue.

@cociboni, that is not the same problem being reported in this issue.

It seems like you’re not really aware of what’s happening, but all day long my sessions have been experiencing exactly the reconnecting issue that others are talking about. Every five minutes there was a reconnecting error, and eventually everything ended up in this state.

Now none of the chats respond at all. I can’t stop them, and when I start a new session it goes into reconnecting again and then falls into the same bug loop.

Instead of replying to each message like this, it would be much more helpful if you tried to actually assist with the issue.

Thank you.

Yes I have the same issue.
request id: 2f94b8a7-9694-4423-b073-5765c06c7f71;
and here is my thread id: 019cc9d3-8034-78d3-b1df-7f24f13f2061

I dont think this is a network problem from our side. I am not on a vpn, proxy or anything. I talked with friends in 5 countries, they all having the same issue.

I am networking engineer.

<img width="472" height="270" alt="Image" src="https://github.com/user-attachments/assets/6cee82ac-dbbd-477d-8e3c-6a1a7c6294e3" />

I'm having the same problem and I don't use a VPN, proxy restrictions, etc... It was working perfectly until yesterday.

I wasn't having problems before today, but I am now. 5-seat business plan using codex CLI. It feels like they are messing with the priority account traffic in order to accommodate lower tiered users because of high traffic. Hope that's not the case. Looks like the online help assistant is overrun too 😆

019cd448-72f7-7ce1-a41c-8fbc6cefb879

same problem here: 019c9f6a-2dfe-7610-a3d5-7891d191e01b

We're seeing very high utilization of Codex today which is leading to long response times and disconnects. We've been working to mitigate the problem. Apologies for any disruption this has caused.

For me, this problem resulted in using up my 5-hour limit 10 times faster and using up 1/3 of my weekly limit.

Things were fine earlier today but tonight? It's this for everything. Pretty frustrating.

<img width="818" height="1200" alt="Image" src="https://github.com/user-attachments/assets/550940fa-8faf-4536-a3a2-237bfcdfb7e6" />

@etraut-openai, you said

Most of these that we've investigated have turned out to be network connectivity issues (flaky networks, VPNs, proxies, firewalls, etc.).

We're looking at implementing a diagnostic utility to help users self-diagnose their networking problems.
I may be able to help you diagnose if you use /feedback to upload your logs and then post your thread ID here.

No, I have no VPN's, firewalls or network connectivity issues at all, and I have two devices, one on the newest version of codex CLI and one on the previous version. The issue only started happening on the device that I upgraded CLI to v0.113.0.

same issue 019cdb8f-181e-70d3-baf7-3651e84679f9

Same issue
019c9fe3-fd85-75c1-96ac-1e060870c254

Same issue. I used 5.4 yesterday and would hit the "Reconnecting" bug every 3-5 outputs on average. As you can see from the screenshot below, my usage almost 3-4x'd a normal day's worth of usage.

The rest of my engineering org also faced similar issues where their weekly usage tanked by 40% even though they never hit their 5h usage limits.

We also saw visible degradation of quality with 5.4. We have advised our team to switch back to 5.3, but we've hit our weekly usage limits twice in as many days.

Can we please get our organization's usage replenished? It's cost us quite a bit of money for a degraded experience and we've wasted time and effort.

<img width="938" height="698" alt="Image" src="https://github.com/user-attachments/assets/c5504767-6e1a-4217-830f-b76bc9ca81cb" />

<img width="430" height="278" alt="Image" src="https://github.com/user-attachments/assets/b0a40c21-674e-4a37-a153-a00a1d874be2" />

still same!

My codex app is of a 12nd Feb version and everything is fine.
Jetbrain AI Assistant app is the latest version and couldn't work with vpn.

same issue: 019cdc55-9766-7ad1-af55-edfb57304034
Windows, PowerShell 7.5.4, vs code ext., wsl

I’m experiencing the same problem with my ChatGPT Plus subscription when using GPT‑5.4 – Codex is almost unusable. However, there is no issue when I access GPT‑5.4 through OpenRouter. It seems that OpenAI may be deprioritizing Plus users when the model is under high demand.

P.S. I'm on corporate VPN and see "Falling back from WebSockets to HTTPS transport". But with Codex 5.3 there is no Problem. Problem is only with GPT 5.4!

<img width="620" height="401" alt="Image" src="https://github.com/user-attachments/assets/a1c5b967-e618-4f90-83ce-fd1349c620d5" />

In my case, it always failed to connect 5/5 at first, and then started working properly without any reconnecting in the same session. So life is still good if I survive from 5-mins wating at every session start.

It smells like App is connecting something else before picking up the coding tasks, and gives up after failure.

Same issue here.

Same issue here.
l used vpn, look likes awalys reconnecting...

Same isuuse here，using VPN， Plus and Free Subscribe always failed to connect

After a few days of taking a break to wait for my weekly rate limits to reset, I got the Reconnecting issue for the first time. I had to back off, updated the Codex App, and then wait a bit and tried again with the same initial settings and prompt that reconnected earlier: "GPT-5.4: Extra High Reasoning, now it seems to work, but I wasted a couple % of usage rate to get started. Here's my thread id: 54c93b31-22d8-40df-b887-5f1add606336 would appreciate getting at least a couple % top up to compensate for the loss.

Reconnecting... 1/5
Reconnecting... 2/5
Reconnecting... 3/5
Reconnecting... 4/5
Reconnecting... 5/5
stream disconnected before completion: error sending request for url (https://chatgpt.com/backend-api/codex/responses)

Same isuuse here

No good for me today (Australia)
Reconnecting... 1/5
Reconnecting... 2/5
Reconnecting... 3/5
Reconnecting... 4/5
Reconnecting... 5/5
We're currently experiencing high demand, which may cause temporary errors.

Chinese:
我是windows 系统，我使用v2ray工具开了代理（VPN），也有一样的网络问题，解决方案是:将代理模式从【系统代理】改为【隧道（Tun）】。
因为windows的【系统代理】是应用层代理，部分流量走不了代理，而【隧道】代理是模拟网卡，是更底层。

mac os 本人不熟悉，猜测应该有类似的方案。

English:

I’m using Windows, and I had the same network issue when using the v2ray tool with a proxy (VPN) enabled. The solution was to change the proxy mode from “System Proxy” to “Tunnel (TUN)”.

This is because “System Proxy” on Windows works at the application layer, so some traffic cannot go through the proxy. In contrast, “Tunnel” mode simulates a network adapter and works at a lower level.

I’m not very familiar with macOS, but I guess there may be a similar solution there as well.

I waited 4 days for my weekly limit to reset, only to encounter this exact same issue.

Codex Desktop 频繁 Reconnecting 的根本原因是：

• OpenAI 访问必须走代理
• Codex 没有稳定命中代理
• FlClash 没有 OpenAI 专用强制代理规则

解决方案是：

• 用用户级环境变量固定 Codex 走 127.0.0.1:7890
• 用 FlClash 规则把 openai.com/chatgpt.com/oaistatic.com/oaiusercontent.com 强制送入代理组
• 重启 FlClash 和 Codex，使配置生效

just a quick note: I contacted support, explained the situation, linked this github thread, and mentioned that many users are experiencing the same issue. We exchanged 3 emails, and in every single one, they asked for 7 different documents and logs. I told them this bug is draining our limits. After providing 21 different documents in total, their latest reply asked for 7 more things. That's when I realized it's completely pointless.

don't waste your time contacting support to get your limits back; they don't care at all. They still think the issue is on our end and keep giving 'have you tried turning it off and on again' type of responses. It’s honestly just frustrating :)

what i ended up doing - just remove .codex folder and restart the app, and it works again, VS Code, App and CLI, no issues...

same to me, both Codex App and VS Code:

<img width="328" height="351" alt="Image" src="https://github.com/user-attachments/assets/c3542a3e-38b4-4ea2-b1c6-245abfc5cec3" />

I paid 20$ and only get this, claude code and gemini all work well in same PC.

The increased rate of reconnects was related to capacity issues last week. We have since mitigated this problem, so I'm going to close and lock this issue.

Many reconnect messages that are reported to us turn out to be client-side network configuration issues or flaky network connections.