--json doesn't output reasoning items when using API Key auth
Open 💬 7 comments Opened Feb 5, 2026 by terzioglub
💡 Likely answer: A maintainer (etraut-openai, contributor)
responded on this thread — see the highlighted reply below.
What version of Codex CLI is running?
CLI 0.94.0
What subscription do you have?
Business
Which model were you using?
gpt-5.2-codex
What platform is your computer?
_No response_
What terminal emulator and version are you using (if applicable)?
_No response_
What issue are you seeing?
Body
When running Codex CLI with --json, no "type":"reasoning" items are included in the JSONL output.
Only command_execution and agent_message events appear. Reasoning output used to be present and still appears for other users on the same version.
This seems account/org dependent. If intentional, please document or provide a flag to enable it.
What steps can reproduce the bug?
Repro:
npm install -g @openai/codex@0.94.0
codex exec --json "analyze this repo"
Expected: JSONL includes reasoning items
Actual: No reasoning items emitted
What is the expected behavior?
_No response_
Additional information
_No response_
7 Comments
Found the cause: reasoning output depends on how Codex is authenticated.
With web login:
codex login
→ --json includes "type":"reasoning" items.
With API key:
export OPENAI_API_KEY=...
→ No reasoning items are emitted.
So reasoning only appears with web auth, not API key auth.
Is there a way to enable reasoning output when using API key authentication as well? If this is intentional, could this be documented or exposed via a flag/config?
Thanks for reporting.
Looks like this isn't a regression, but it is a bug — not intended behavior.
Thanks for confirming.
Happy to help if there’s anything that can be done on the CLI side, but from what I can tell this looks like a backend/auth behavior difference (web vs API key), since the CLI never receives reasoning items in API-key mode.
Let me know if there’s anything I can test or contribute on the client side.
Hey @etraut-openai have you had a chance to tackle this issue ? this is causing some trouble for us for the past few weeks
Hello Team
I was wondering if there is any updates on this issue and I would be also glad to tackle this issue , if it can be fixed on the open source side of things , do you accept contributions ?
cc @etraut-openai
hi folks, any updates on this?
hey team this issue has been open for the past three months , is there anyway someone picks this up soon ?