Codex CLI v0.63.0 always interrupts conversation and hangs on basic prompts (Pop!_OS)

What version of Codex is running?

codex --version OpenAI Codex (v0.63.0)

What subscription do you have?

ChatGPT Plus + Project API Key (with active billing credits)

Which model were you using?

gpt-5.1-codex-max (reasoning: medium)

What platform is your computer?

Pop!_OS 22.04 (Ubuntu-based) Linux x86_64 (node v20.19.6, npm v11.6.4)

What issue are you seeing?

Title: Codex CLI v0.63.0 interrupts every prompt and produces no output

Version:
OpenAI Codex (v0.63.0)

Subscription:
ChatGPT Plus + Project API Key

Model:
gpt-5.1-codex-max

Platform:
Pop!_OS 22.04 (Ubuntu)
Linux x86_64
node v20.19.6
npm v11.6.4

Issue Description:
Codex CLI authenticates successfully but cannot respond to any prompt. Every message immediately fails with:

■ Conversation interrupted - tell the model what to do differently. Something went wrong?

This happens with simple inputs like "hello". Codex never produces output. /status shows a valid session, valid API key, and zero tokens used. Other OpenAI APIs work fine on this machine.

Example:
$ codex

hello

■ Conversation interrupted - tell the model what to do differently. Something went wrong?

$ codex /status
Model: gpt-5.1-codex-max
API key configured
Session: 019acd8f-2841-7a50-a804-99571fb9ea4a
Token usage: 0 total

Steps to Reproduce:

1. Install Codex with: npm i -g @openai/codex
2. Run: codex
3. Authenticate with ChatGPT login or project API key
4. Type any prompt (e.g., "hello")
5. Codex immediately fails with the above error

Expected Behavior:
Codex should respond with completions or perform actions. Instead, every prompt is interrupted and no output is generated.

Additional Info:

• Issue occurs 100% of the time
• Reinstalling does not fix it
• Using ChatGPT login or project key makes no difference
• All other OpenAI API calls work normally

What steps can reproduce the bug?

1. Install Codex CLI:

npm i -g @openai/codex

1. Launch Codex:

codex

1. Authenticate using either:

• ChatGPT Plus login, or
• Project API key

1. After Codex loads, type any prompt:

hello

1. Codex immediately fails with:

■ Conversation interrupted - tell the model what to do differently. Something went wrong?

This happens 100% of the time, even in an empty directory and even with different authentication methods.

Session ID example:
019acd8f-2841-7a50-a804-99571fb9ea4a

Token usage:
0 total (0 input / 0 output)

Context window usage:
0 used / 272K available

Codex does not execute any tasks and does not generate any output.

What is the expected behavior?

Codex CLI should accept a prompt and generate a normal model response. For simple inputs like "hello," Codex should return a short completion instead of interrupting the conversation. For coding-related tasks, Codex should analyze the directory, read files, propose edits, or execute agent actions without failing immediately.

In general, expected behavior is:

• Codex processes the input prompt
• Codex produces a model-generated response or action plan
• Token usage increases from 0
• The session continues normally without interruption

Instead, Codex stops instantly with an error and provides no output.

Additional information

Codex CLI worked previously on this same machine but began failing recently without any configuration changes. The issue persists across reinstalls, different directories, different authentication methods, DNS resets, and system restarts. Other OpenAI API tools (curl requests and the official OpenAI CLI) work correctly with the same project API key. This failure appears isolated to the Codex CLI executable itself.

If helpful: DNS resolution for cli.openai.com returns NXDOMAIN on all major resolvers (Cloudflare, Google), while other OpenAI domains resolve correctly. This may indicate an upstream infrastructure issue affecting the CLI.