Bug type
Behavior bug (incorrect output/state without crash)
Beta release blocker
No
Summary
On OpenClaw 2026.5.3, a session using openai-codex/gpt-5.5 with active openai-codex OAuth no longer shows Codex 5h/weekly usage in /status / session_status, and openclaw status --usage reports Codex: Unsupported provider.
Steps to reproduce
- Run OpenClaw 2026.5.3 with an
openai-codex OAuth profile configured.
- Use a session whose active model is
openai-codex/gpt-5.5.
- Run
/status or inspect session_status.
- Run
openclaw status --usage.
- Observe that the status card shows model/auth but no Codex usage windows, while CLI usage reports
Codex: Unsupported provider.
Expected behavior
For an active openai-codex OAuth session, /status / session_status should show Codex quota windows such as 5h and Week/weekly usage, or at minimum openclaw status --usage should return Codex usage/quota data instead of Unsupported provider.
This is also consistent with the documented openclaw status --usage behaviour: provider usage windows should be printed as normalized % left values.
Actual behavior
session_status shows the active Codex model and OAuth profile, but no usage/quota line:
🦞 OpenClaw 2026.5.3 (002b360)
🧠 Model: openai-codex/gpt-5.5 · 🔑 oauth (openai-codex:<redacted-email>)
🧮 Tokens: 18k in / 69 out
🗄️ Cache: 17% hit · 3.6k cached, 0 new
📚 Context: 21k/200k (11%)
...
Running openclaw status --usage reports:
Usage
Usage:
Claude: HTTP 403: OAuth token does not meet scope requirement user:profile
Codex: Unsupported provider
The Codex usage line is therefore absent from /status / session_status.
OpenClaw version
OpenClaw 2026.5.3 (002b360)
Operating system
Ubuntu 24.04 arm64 / Linux 6.17.0-1011-oracle
Install method
npm global
/usr/lib
└── openclaw@2026.5.3
Model
openai-codex/gpt-5.5
Provider / routing chain
OpenClaw -> openai-codex OAuth / ChatGPT Codex quota endpoint
Additional provider/model setup details
- Default model:
openai-codex/gpt-5.5
- Configured aliases:
codex -> openai-codex/gpt-5.5
codex54 -> openai-codex/gpt-5.4
- Auth overview shows
openai-codex OAuth profiles available and valid.
- The active session status shows
🔑 oauth (openai-codex:<redacted-email>).
- This does not appear to be an alias-resolution issue: the selected provider/model and auth profile are correctly displayed as
openai-codex.
Logs, screenshots, and evidence
$ openclaw --version
OpenClaw 2026.5.3 (002b360)
$ command -v openclaw
/usr/bin/openclaw
$ node --version
v22.22.0
$ uname -a
Linux polymarket-mc 6.17.0-1011-oracle #11~24.04.1-Ubuntu SMP Fri Apr 10 02:09:09 UTC 2026 aarch64 aarch64 aarch64 GNU/Linux
$ npm list -g openclaw --depth=0
/usr/lib
└── openclaw@2026.5.3
$ openclaw models
Default : openai-codex/gpt-5.5
Fallbacks (1) : google/gemini-2.5-flash
Aliases (2) : codex -> openai-codex/gpt-5.5, codex54 -> openai-codex/gpt-5.4
Configured models (2): openai-codex/gpt-5.5, openai-codex/gpt-5.4
OAuth/token status
- openai-codex
- openai-codex:default (<redacted-email>) ok expires in 7d
- openai-codex:<redacted-email> (<redacted-email>) ok expires in 43h
$ openclaw status --usage
Usage
Usage:
Claude: HTTP 403: OAuth token does not meet scope requirement user:profile
Codex: Unsupported provider
Relevant local docs state:
openclaw status --usage
--usage prints normalized provider usage windows as X% left.
Related existing GitHub items found during triage:
Impact and severity
Affected: Users relying on OpenClaw /status, session_status, or openclaw status --usage to monitor ChatGPT/Codex OAuth usage windows.
Severity: Medium. Core model execution still works, but quota visibility is missing, making it harder to manage Codex usage and avoid unexpected quota exhaustion.
Frequency: Reproduced on the current 2026.5.3 install when running openclaw status --usage.
Consequence: Operators lose the 5h/weekly Codex usage gauge previously visible in status output and must manually infer remaining quota elsewhere.
Additional information
The active model/auth display appears correct, so this does not look like a simple alias issue. The failure appears specifically in the Codex provider usage/quota path: status resolves openai-codex auth successfully, but the usage report falls through to Codex: Unsupported provider.
Bug type
Behavior bug (incorrect output/state without crash)
Beta release blocker
No
Summary
On OpenClaw 2026.5.3, a session using
openai-codex/gpt-5.5with activeopenai-codexOAuth no longer shows Codex 5h/weekly usage in/status/session_status, andopenclaw status --usagereportsCodex: Unsupported provider.Steps to reproduce
openai-codexOAuth profile configured.openai-codex/gpt-5.5./statusor inspectsession_status.openclaw status --usage.Codex: Unsupported provider.Expected behavior
For an active
openai-codexOAuth session,/status/session_statusshould show Codex quota windows such as 5h and Week/weekly usage, or at minimumopenclaw status --usageshould return Codex usage/quota data instead ofUnsupported provider.This is also consistent with the documented
openclaw status --usagebehaviour: provider usage windows should be printed as normalized% leftvalues.Actual behavior
session_statusshows the active Codex model and OAuth profile, but no usage/quota line:Running
openclaw status --usagereports:The Codex usage line is therefore absent from
/status/session_status.OpenClaw version
OpenClaw 2026.5.3 (002b360)
Operating system
Ubuntu 24.04 arm64 / Linux 6.17.0-1011-oracle
Install method
npm global
Model
openai-codex/gpt-5.5
Provider / routing chain
OpenClaw -> openai-codex OAuth / ChatGPT Codex quota endpoint
Additional provider/model setup details
openai-codex/gpt-5.5codex -> openai-codex/gpt-5.5codex54 -> openai-codex/gpt-5.4openai-codexOAuth profiles available and valid.🔑 oauth (openai-codex:<redacted-email>).openai-codex.Logs, screenshots, and evidence
Relevant local docs state:
Related existing GitHub items found during triage:
session_statususage data is unstable / usage fields intermittently missingstatus --usage --jsonhanging/failing pathImpact and severity
Affected: Users relying on OpenClaw
/status,session_status, oropenclaw status --usageto monitor ChatGPT/Codex OAuth usage windows.Severity: Medium. Core model execution still works, but quota visibility is missing, making it harder to manage Codex usage and avoid unexpected quota exhaustion.
Frequency: Reproduced on the current 2026.5.3 install when running
openclaw status --usage.Consequence: Operators lose the 5h/weekly Codex usage gauge previously visible in status output and must manually infer remaining quota elsewhere.
Additional information
The active model/auth display appears correct, so this does not look like a simple alias issue. The failure appears specifically in the Codex provider usage/quota path: status resolves
openai-codexauth successfully, but the usage report falls through toCodex: Unsupported provider.