ClawRouter continues consuming credits despite primary model set to Codex

[crazy2705]

ClawRouter continues consuming credits despite primary model set to Codex

Problem Description

After configuring openai-codex/gpt-5.3-codex as the primary model in the configuration file, ClawRouter continues to intercept and consume credits for every request, completely ignoring the model setting.

Documented Credit Loss

Time | Remaining Credit | Loss -- | -- | -- 11:42 | $5.45 | - 11:56 | $4.02 | -$1.43 12:01 | $3.78 | -$0.24 12:07 | $3.19 | -$0.59 Total |   | -$2.26+

Current Configuration (agents.defaults.model)

{
  "primary": "codex",
  "fallbacks": ["auto", "openrouter/auto", "gemini-2.5-flash"]
}

Expected Behavior

• Requests should use OpenAI Codex (OAuth authentication, included in subscription)
• No ClawRouter credit consumption for normal chat requests

Actual Behavior

• [ClawRouter] Received model: "auto" in logs despite primary: "codex"
• Usage of paid models (xai/grok-4-fast-reasoning, moonshot/kimi-k2.5)
• Cost per call: ~$0.0033
• Current session shows model: "auto" instead of "codex"

Relevant Logs

[ClawRouter] Received model: "auto" -> normalized: "auto", isAuto: true
[ClawRouter] Tools detected (23), forcing agentic mode
[ClawRouter] Trying model 1/2: moonshot/kimi-k2.5
[ClawRouter] Success with model: moonshot/kimi-k2.5
[plugins] [REASONING] moonshot/kimi-k2.5 $0.0033 (saved 99%)

Steps to Reproduce

1. Configure primary: "codex" in agents.defaults.model
2. Restart OpenClaw (SIGUSR1)
3. Verify session_status shows model: "openai-codex/gpt-5.3-codex"
4. Execute any request
5. Observe logs still show [ClawRouter] Received model: "auto"

Attempted Solutions

• Changed primary from blockrun/auto to codex
• Removed auto from fallbacks
• Set session_status with model: "codex" for current session
• Restarted OpenClaw multiple times
• Completely disabled ClawRouter plugin - not attempted for fear of losing functionality

Environment

• OpenClaw Version: 2026.2.9 (33c75cb)
• OS: Linux 6.12.63+deb13-amd64 (x64)
• Node: v22.22.0

Additional Context

The issue appears to be that even when the session status shows the correct model (openai-codex/gpt-5.3-codex), the underlying tool calls or agent processing still routes through ClawRouter with model: "auto". This suggests the model override is not being properly propagated to all subsystems, particularly when tools are involved ("Tools detected (23), forcing agentic mode").

Possible Root Causes

1. Tool calls force agentic mode which overrides user model preference
2. Session model override doesn't propagate to ClawRouter plugin
3. auto alias in fallbacks causes ClawRouter to activate despite primary being set
4. Configuration patch applies to defaults but active sessions retain cached model

Request

Please investigate why ClawRouter continues to intercept requests when a specific non-ClawRouter primary model is configured. The expected behavior is that explicit model selection should bypass ClawRouter entirely unless explicitly requested via /model auto.

[1bcMax]

Let me take a look. Can I have your wallet information?

[1bcMax]

Investigation Results

After examining OpenClaw's source code, I found the root cause:

Why ClawRouter Receives model: auto

Your session_status shows openai-codex/gpt-5.3-codex, but when the actual API request fails, OpenClaw's fallback chain kicks in:

Primary: openai-codex/gpt-5.3-codex → FAILS
Fallback 1: "auto" → ClawRouter intercepts → CREDITS USED

Potential Causes

1. Invalid model name - OpenClaw's valid codex models are:

   • openai-codex/gpt-5.2-codex
   • openai-codex/gpt-5.1-codex

   gpt-5.3-codex may not exist, causing request failure.

2. Authentication issue - Your openai-codex OAuth token may have expired or hit quota limits.

The Fix

1. Use a valid model name:

   {
     "primary": "openai-codex/gpt-5.2-codex",
     "fallbacks": ["google/gemini-2.5-flash", "deepseek/deepseek-chat"]
   }

2. Remove "auto" from fallbacks - this triggers ClawRouter smart routing

3. Alternative cheap fallbacks (BlockRun models):

   • google/gemini-2.5-flash - $0.15/$0.60 per 1M tokens
   • deepseek/deepseek-chat - $0.28/$0.42 per 1M tokens
   • nvidia/gpt-oss-120b - FREE
   • free - routes to nvidia only, no cost

How to Verify

Run /models in OpenClaw to check which models are available for your authenticated providers.

---

This is not a ClawRouter bug - ClawRouter correctly processes auto requests when they arrive. The issue is that your primary model fails silently and falls back to auto.

[crazy2705]

I cannot confirm the analysis because the selected model was working and it's reported in various official soures
https://docs.openclaw.ai/providers/openai
{
agents: { defaults: { model: { primary: "openai-codex/gpt-5.3-codex" } } },
}

My wallet is 0xdC10bb5B851D46A42AddF2dDC8bc02BC95fDc315
I spent 65$ in less than 4 days.

Thanks for your support, in case let me know.

[1bcMax]

@crazy2705 can you dm me on tg @bc1max or join our group https://t.me/+mroQv4-4hGgzOGUx

[1bcMax]

I try to add anthropic api and blockrun side by side. and try to work on it , but i didnt see BlockRun charge the same time. The bill is clearly on anthropic api side..