#Confusing harness path for GPT model

1 messages · Page 1 of 1 (latest)

noble lake
#

I've updated to the latest openclaw version and have been trying to get the GPT5.5 model to run via the codex harness instead of the Pi harness. No matter what I or codex-cli does I just can't get it to hook up to the codex harness. I've burnt though so many codex token trying to get it sorted out.

This is a sanitised extract from my openclaw.config

      "profiles": {
        "minimax:global": {
          "provider": "minimax",
          "mode": "api_key"
        },
        "minimax:cn": {
          "provider": "minimax",
          "mode": "api_key"
        },
        "openai-codex:<REDACTED_EMAIL>": {
          "provider": "openai-codex",
          "mode": "oauth"
        }
      }
    },
    "agents": {
      "defaults": {
        "model": {
          "primary": "minimax/MiniMax-M2.7",
          "fallbacks": []
        },
        "models": {
          "minimax/MiniMax-M2.7": {
            "alias": "Minimax"
          },
          "openai-codex/gpt-5.5": {}
        },
        "agentRuntime": {
          "id": "codex",
          "fallback": "pi"
        }
      },```
#

``"plugins": {
"allow": [
"codex",
"memory-core",
"memory-wiki",
"minimax",
"nvidia",
"openai",
"telegram"
],
"entries": {
"telegram": {
"enabled": true
},
"minimax": {
"enabled": true
},
"openai": {
"enabled": true
},
"memory-core": {
"enabled": true,
"config": {
"dreaming": {
"enabled": true
}
}
},
"memory-wiki": {
"enabled": true
},
"nvidia": {
"enabled": true
},
"codex": {
"enabled": true```

keen sundial
#

I had to remove this entire section to make it work for mine.

"agentRuntime": {
"id": "codex",
“fallback”: “pi”
},

Once I did that it seemed to start working correctly with oauth.