Fix: 9.98s Verification Timeout
Why "curl" Lies to You
When you test an endpoint with curl, you are measuring raw TCP connectivity. However, OpenClaw's Preflight Layer performs a full /v1/models request to validate your API key and model availability. Many API aggregators or local servers (like LM Studio) have high latency on this specific endpoint, often exceeding OpenClaw's internal 10,000ms safety cutoff.
If your logs show verification timed out after ~9.98s, OpenClaw is killing the connection before your provider can reply with its model list.
Next Step
Fix now, then reduce repeat incidents
If this issue keeps coming back, validate your setup in Doctor first, then harden your config.
The Diagnosis
Confirm if your provider's model list response is the bottleneck. Time the actual application-level response:
time curl -H 'Authorization: Bearer YOUR_KEY' https://your-provider.com/v1/models
If the "real" time is over 8 seconds, OpenClaw will almost certainly fail to verify.
Solution 1: Bypass Discovery (Best Fix)
Instead of letting OpenClaw ask the provider "what models do you have?", tell it exactly which models to use. This skips the 10s preflight check entirely.
// In openclaw.json
"providers": {
"my-proxy": {
"module": "openai",
"config": {
"baseURL": "https://api.proxy.com/v1",
"models": ["gpt-4o", "claude-3-sonnet"], // Explicit list
"skipVerification": true // Skip preflight
}
}
}Solution 2: Increase Preflight Timeout
If you prefer dynamic discovery, you can increase the internal discovery timeout in your global settings.
Pro Tip: Local Servers
For local tools like Ollama or LM Studio, timeouts are often due to the model being "swapped into memory" during the first request. Pre-load your model before starting the OpenClaw gateway to ensure sub-second response times.
Did this guide solve your problem?