Issue Station
Sign In
Tools hereToolbox

openclaw - 💡(How to fix) Fix [Bug]: Codex harness rejects openai/gpt-5.5 after 2026.5.27 update

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…

After attempting to update to OpenClaw 2026.5.27, all agent turns using the Codex harness failed before reply with:

Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex).

The failure also reproduced after trying the openai-codex/gpt-5.5 model ref:

Requested agent harness "codex" does not support openai-codex/gpt-5.5 (provider is not one of: codex).

This effectively made every model unavailable when non-Codex fallbacks were in billing/cooldown, even though Codex OAuth auth was present and later worked after rollback.

Error Message

[diagnostic] lane task error: lane=main durationMs=134 error="Error: Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex)." [diagnostic] lane task error: lane=session:<redacted> durationMs=137 error="Error: Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex)." [model-fallback/decision] decision=candidate_failed requested=zai/glm-5-turbo candidate=openai/gpt-5.5 reason=unknown next=zai/glm-5.1 detail=Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex).

Root Cause

High operational impact: after updating to 2026.5.27, no model could answer in this environment until rollback, because the only healthy provider was OpenAI Codex OAuth and the Codex harness rejected the model refs.

Fix Action

Workaround

Rollback to 2026.5.22 restored functionality in this environment. Current status after rollback:

OpenClaw 2026.5.22 (a374c3a)
openai/gpt-5.5 ... Auth yes ... default,configured
Runtime auth: openai via codex ... status=usable

Code Example

Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex).

---

Requested agent harness "codex" does not support openai-codex/gpt-5.5 (provider is not one of: codex).

---

[model-fallback/decision] decision=candidate_failed requested=zai/glm-5-turbo candidate=zai/glm-5-turbo reason=billing next=openai/gpt-5.5 detail=429 Insufficient balance or no resource package. Please recharge.

---

[diagnostic] lane task error: lane=main durationMs=134 error="Error: Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex)."
[diagnostic] lane task error: lane=session:<redacted> durationMs=137 error="Error: Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex)."
[model-fallback/decision] decision=candidate_failed requested=zai/glm-5-turbo candidate=openai/gpt-5.5 reason=unknown next=zai/glm-5.1 detail=Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex).

---

Embedded agent failed before reply: All models failed (3):
- zai/glm-5-turbo: 429 Insufficient balance or no resource package. Please recharge. (billing)
- openai/gpt-5.5: Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex). (unknown)
- zai/glm-5.1: No available auth profile for zai (all in cooldown or unavailable). (billing)

---

[gateway] agent model: openai-codex/gpt-5.5 (thinking=off, fast=off)

---

[diagnostic] lane task error: lane=main durationMs=77 error="Error: Requested agent harness "codex" does not support openai-codex/gpt-5.5 (provider is not one of: codex)."
[model-fallback/decision] decision=candidate_failed requested=openai-codex/gpt-5.5 candidate=openai-codex/gpt-5.5 reason=unknown next=zai/glm-5.1 detail=Requested agent harness "codex" does not support openai-codex/gpt-5.5 (provider is not one of: codex).
Embedded agent failed before reply: All models failed (3): openai-codex/gpt-5.5: Requested agent harness "codex" does not support openai-codex/gpt-5.5 (provider is not one of: codex). (unknown) | ...

---

Your OpenClaw config was written by version 2026.5.27, but this command is running 2026.5.22.

---

Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex).

---

Requested agent harness "codex" does not support openai-codex/gpt-5.5 (provider is not one of: codex).

---

OpenClaw 2026.5.22 (a374c3a)
openai/gpt-5.5 ... Auth yes ... default,configured
Runtime auth: openai via codex ... status=usable
RAW_BUFFERClick to expand / collapse

Summary

After attempting to update to OpenClaw 2026.5.27, all agent turns using the Codex harness failed before reply with:

Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex).

The failure also reproduced after trying the openai-codex/gpt-5.5 model ref:

Requested agent harness "codex" does not support openai-codex/gpt-5.5 (provider is not one of: codex).

This effectively made every model unavailable when non-Codex fallbacks were in billing/cooldown, even though Codex OAuth auth was present and later worked after rollback.

Environment

Sanitized from local logs/status:

  • OpenClaw attempted update: 2026.5.27
  • Rollback/current working version: 2026.5.22 (a374c3a)
  • Gateway plugins loaded during the failing window included: browser, codex, discord, exa, lossless-claw, memory-core, memory-wiki, telegram
  • Runtime/harness involved: Codex harness / OpenAI Codex runtime
  • Auth state: OpenAI Codex OAuth profiles were present; after rollback, openai/gpt-5.5 is usable through Codex OAuth
  • Non-Codex fallback provider was unavailable due to billing/cooldown, which exposed this as total model failure

No local paths, chat IDs, session IDs, tokens, or account emails are included here.

Relevant sanitized log sequence

  1. Fallback starts from a non-Codex model that is unavailable due to billing:
[model-fallback/decision] decision=candidate_failed requested=zai/glm-5-turbo candidate=zai/glm-5-turbo reason=billing next=openai/gpt-5.5 detail=429 Insufficient balance or no resource package. Please recharge.
  1. The fallback to canonical OpenAI model fails inside the Codex harness provider check:
[diagnostic] lane task error: lane=main durationMs=134 error="Error: Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex)."
[diagnostic] lane task error: lane=session:<redacted> durationMs=137 error="Error: Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex)."
[model-fallback/decision] decision=candidate_failed requested=zai/glm-5-turbo candidate=openai/gpt-5.5 reason=unknown next=zai/glm-5.1 detail=Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex).
  1. With the remaining provider in cooldown, the whole turn fails before reply:
Embedded agent failed before reply: All models failed (3):
- zai/glm-5-turbo: 429 Insufficient balance or no resource package. Please recharge. (billing)
- openai/gpt-5.5: Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex). (unknown)
- zai/glm-5.1: No available auth profile for zai (all in cooldown or unavailable). (billing)
  1. After changing the configured model namespace, the gateway logged:
[gateway] agent model: openai-codex/gpt-5.5 (thinking=off, fast=off)

but turns still failed similarly:

[diagnostic] lane task error: lane=main durationMs=77 error="Error: Requested agent harness "codex" does not support openai-codex/gpt-5.5 (provider is not one of: codex)."
[model-fallback/decision] decision=candidate_failed requested=openai-codex/gpt-5.5 candidate=openai-codex/gpt-5.5 reason=unknown next=zai/glm-5.1 detail=Requested agent harness "codex" does not support openai-codex/gpt-5.5 (provider is not one of: codex).
Embedded agent failed before reply: All models failed (3): openai-codex/gpt-5.5: Requested agent harness "codex" does not support openai-codex/gpt-5.5 (provider is not one of: codex). (unknown) | ...
  1. During rollback/restart, logs showed a version/config mismatch warning:
Your OpenClaw config was written by version 2026.5.27, but this command is running 2026.5.22.

After rollback to 2026.5.22, the model catalog/auth status is healthy again and openai/gpt-5.5 works through Codex OAuth.

Expected behavior

The Codex harness should accept the model reference format OpenClaw uses for Codex OAuth-backed OpenAI models, or normalize/route it consistently before the harness provider check.

At minimum, one of these should work reliably:

  • openai/gpt-5.5 with Codex OAuth / Codex runtime
  • openai-codex/gpt-5.5
  • codex/gpt-5.5

The current failure suggests the harness/provider namespace validation is rejecting both canonical openai/* and attempted openai-codex/* refs, while reporting that provider is not codex.

Actual behavior

Agent turns fail before reply with:

Requested agent harness "codex" does not support openai/gpt-5.5 (provider is not one of: codex).

or:

Requested agent harness "codex" does not support openai-codex/gpt-5.5 (provider is not one of: codex).

Fallback then exhausts all candidates when other providers are unavailable.

Impact

High operational impact: after updating to 2026.5.27, no model could answer in this environment until rollback, because the only healthy provider was OpenAI Codex OAuth and the Codex harness rejected the model refs.

Workaround

Rollback to 2026.5.22 restored functionality in this environment. Current status after rollback:

OpenClaw 2026.5.22 (a374c3a)
openai/gpt-5.5 ... Auth yes ... default,configured
Runtime auth: openai via codex ... status=usable

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…

FAQ

Expected behavior

The Codex harness should accept the model reference format OpenClaw uses for Codex OAuth-backed OpenAI models, or normalize/route it consistently before the harness provider check.

At minimum, one of these should work reliably:

  • openai/gpt-5.5 with Codex OAuth / Codex runtime
  • openai-codex/gpt-5.5
  • codex/gpt-5.5

The current failure suggests the harness/provider namespace validation is rejecting both canonical openai/* and attempted openai-codex/* refs, while reporting that provider is not codex.

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING