OpenClaw is an open-source personal AI assistant. Cobble plugs in as a custom provider — one config block, plus one gotcha (the model allowlist) that this guide covers so you don't hit it.
Works best with: Agent Runner. OpenClaw runs continuously, so the pinned slot and deep request window matter more than raw model size.
1. Add Cobble as a provider
In your OpenClaw config, add Cobble under models.providers:
{
"models": {
"providers": {
"cobble": {
"baseUrl": "https://api.cobble.network/v1",
"apiKey": "sk-cobble-...",
"api": "openai-completions",
"models": [
{
"id": "qwen/qwen3.5-122b-a10b",
"name": "Qwen3.5 122B A10B",
"contextWindow": 128000
}
]
}
}
}
}2. Allowlist the model — don't skip this
OpenClaw rejects any model not listed in agents.defaults.models, using the format provider-name/model-id:
{
"agents": {
"defaults": {
"models": ["cobble/qwen/qwen3.5-122b-a10b"]
}
}
}If you skip this step, the provider config looks correct but every request is refused — it's the most common setup failure.
3. Restart and select
Restart OpenClaw, then set your default model to cobble/qwen/qwen3.5-122b-a10b.
Troubleshooting
- "Model not allowed" — step 2. The allowlist entry needs the
cobble/prefix in front of the full model ID. 400errors from the API — make sureapiis"openai-completions". OpenClaw automatically disables OpenAI-only request fields for non-OpenAI hosts, but only when the API type is right.401— key issue; test with curl against/v1/modelsfirst.429overnight — an always-on assistant with heartbeats can drain a plan window while you sleep. Either lengthen OpenClaw's polling interval or set a small wallet spend cap as a buffer.
Plan notes
OpenClaw is the most usage-hungry tool we support — it's an assistant that never logs off. Agent Runner's 750 requests per rolling 5-hour window comfortably covers normal conversational use, but aggressive heartbeat/cron configurations are the classic way to hit 429s. Tune the intervals before upgrading plans.
