nothing happening in response to me asking it to do things

yup this happens to me as well

Same. This is blowing up multiple active workstreams

Same here. Using Windows 11, Codex experimental sandbox on Powershell.

Same here on mac

They seem to have an outage again. Thread 019a555c-f8f9-7183-83fa-7c482c4646c0

yeah, seems like an outage

Same, Ubuntu 24.04.03 definitely an outage

Can confirm, the model will start working and then abruptly stop (I get the turn done notification triggered in my script as well). I've tried starting new sessions, switching models (GPT5/Codex medium/high), but everything leads to the model immediately abandoning attempts.

Arch Linux x86_64
Linux 6.17.7-zen1-1-zen
KDE Plasma 6.5.1 (Wayland, nvidia drivers)
Version: codex-cli 0.55.0

Steps to reproduce

• Ask codex via CLI to complete any task

Expected outcome

• Codex completes work and notifies when done

Actual outcome

• Codex starts working on the task and very shortly and abruptly completes its turn without completing the task

Same here, arch. OAI fucked something, doesnt seem to be the CLI, probably the API itself?

Same here!

Same for me

yep. linux mac and windows. @tibo-openai @ae-openai

same for me, 019a5581-f58c-7bb3-8d70-d79b86a69287

I just started having this issue in windows about 5-10 minutes ago. I tried changing the models as well. It works for 4-5 seconds and stops.

yep. session 019a5582-38cc-7ba3-a3b9-28efbb3986b7

same 0.55 mac os, 5 min ago, codex unworkable, session 019a5507-84fa-7c31-a20b-109747f50fb3

Same... happening with session 019a5585-69b6-73f0-93b7-eb83040e0663

and me 019a5586-2ebd-78c1-b8a7-16934c960aba

same as me: 019a5585-ba6b-7140-ae4f-dd58e61353ae

<img src="https://media4.giphy.com/media/v1.Y2lkPWJkM2VhNTdlOG9jaDM2M254ZGhvajAxczhnZGMyODBvNmxycGo1cGlqbG5jcjUzbCZlcD12MV9naWZzX3NlYXJjaCZjdD1n/QCJFiJfnDCdFa84sLG/giphy.gif"/>

having the same issue here

• sessionId: 019a554d-70e5-7023-8882-52ff3eca8fe2
• using codex on my MacBook M3 running Sequoia 15.5
• using package @openai/codex@0.56.0-alpha.2, same behavior on the latest regular @openai/codex when testing as well
• terminal: Ghostty version 1.2.1

Same here...

same

Specifically it seems like it fails when it has to do any tool calls. Just pure chatting works fine

same

same

Same with normal API usage. Here is what my LangGraph agent says:

(agents.crux_generator.node_determine_action:682) node_determine_action
/home/ubuntu/env_langgraph6/lib/python3.10/site-packages/pydantic/main.py:464: UserWarning: Pydantic serializer warnings:
  PydanticSerializationUnexpectedValue(Expected `literal['message']` - serialized value may not be as expected [field_name='type', input_value='shell_call', input_type=str])
  PydanticSerializationUnexpectedValue(Expected `ResponseFileSearchToolCall` - serialized value may not be as expected [field_name='output', input_value=ResponseOutputMessage(id=...5qyoR1Nj5NxqrpyGZZOOYC'), input_type=ResponseOutputMessage])
  PydanticSerializationUnexpectedValue(Expected `ResponseFunctionToolCall` - serialized value may not be as expected [field_name='output', input_value=ResponseOutputMessage(id=...5qyoR1Nj5NxqrpyGZZOOYC'), input_type=ResponseOutputMessage])
  PydanticSerializationUnexpectedValue(Expected `ResponseFunctionWebSearch` - serialized value may not be as expected [field_name='output', input_value=ResponseOutputMessage(id=...5qyoR1Nj5NxqrpyGZZOOYC'), input_type=ResponseOutputMessage])
  PydanticSerializationUnexpectedValue(Expected `ResponseComputerToolCall` - serialized value may not be as expected [field_name='output', input_value=ResponseOutputMessage(id=...5qyoR1Nj5NxqrpyGZZOOYC'), input_type=ResponseOutputMessage])
  PydanticSerializationUnexpectedValue(Expected `ResponseReasoningItem` - serialized value may not be as expected [field_name='output', input_value=ResponseOutputMessage(id=...5qyoR1Nj5NxqrpyGZZOOYC'), input_type=ResponseOutputMessage])
  PydanticSerializationUnexpectedValue(Expected `ImageGenerationCall` - serialized value may not be as expected [field_name='output', input_value=ResponseOutputMessage(id=...5qyoR1Nj5NxqrpyGZZOOYC'), input_type=ResponseOutputMessage])
  PydanticSerializationUnexpectedValue(Expected `ResponseCodeInterpreterToolCall` - serialized value may not be as expected [field_name='output', input_value=ResponseOutputMessage(id=...5qyoR1Nj5NxqrpyGZZOOYC'), input_type=ResponseOutputMessage])
  PydanticSerializationUnexpectedValue(Expected `LocalShellCall` - serialized value may not be as expected [field_name='output', input_value=ResponseOutputMessage(id=...5qyoR1Nj5NxqrpyGZZOOYC'), input_type=ResponseOutputMessage])
  PydanticSerializationUnexpectedValue(Expected `McpCall` - serialized value may not be as expected [field_name='output', input_value=ResponseOutputMessage(id=...5qyoR1Nj5NxqrpyGZZOOYC'), input_type=ResponseOutputMessage])
  PydanticSerializationUnexpectedValue(Expected `McpListTools` - serialized value may not be as expected [field_name='output', input_value=ResponseOutputMessage(id=...5qyoR1Nj5NxqrpyGZZOOYC'), input_type=ResponseOutputMessage])
  PydanticSerializationUnexpectedValue(Expected `McpApprovalRequest` - serialized value may not be as expected [field_name='output', input_value=ResponseOutputMessage(id=...5qyoR1Nj5NxqrpyGZZOOYC'), input_type=ResponseOutputMessage])
  PydanticSerializationUnexpectedValue(Expected `ResponseCustomToolCall` - serialized value may not be as expected [field_name='output', input_value=ResponseOutputMessage(id=...5qyoR1Nj5NxqrpyGZZOOYC'), input_type=ResponseOutputMessage])
  return self.__pydantic_serializer__.to_python(
(agents.crux_generator.node_determine_action:682) node_determine_action
✖ ERROR: (app_configure.aitl_monitor._run_agent:350) [AITL Monitor] Agent launch failed for ticket id=1053: Error code: 400 - {'error': {'message': "Item
'rs_01f68904a5d6aa1000690ba8dd5b288191b0792f81ecd3c0f7' of type 'reasoning' was provided without its required following item.", 'type': 'invalid_request_error', 'param':
'input', 'code': None}}

same with review
\>\> Code review started: current changes <<

<< Code review finished >>

Specifically it seems like it fails when it has to do any tool calls. Just pure chatting works fine

This is the problem, improper response formatting for tool calls coming back from the API right now.. Someone accidentally hit publish without a proper review on the API side potentially. Cause it does seem like just talking to it works ok.

Seems that the model is still live, just the wrapper for CLI and GUI in ide is breaking

same

https://status.openai.com/incidents/01K9ARX310CM6V219R0GF7RNC9

We had a temporary service outage that affected codex. Should be working again now. Let us know if you're still seeing a problem.