Follow OpenClaw installation and run openclaw onboard. Configure a custom provider in ~/.openclaw/openclaw.json (under your user profile on Windows). 
{
  models: {
    mode: "merge",
    providers: {
      infistar: {
        baseUrl: "https://infistar.ai/v1",
        apiKey: "${INFISTAR_API_KEY}",
        api: "openai-completions",
        models: [{ id: "MODEL_ID", name: "MODEL_ID" }]
      }
    }
  },
  agents: { defaults: { model: { primary: "infistar/MODEL_ID" } } }
}
Set INFISTAR_API_KEY in the environment of the process running OpenClaw. Replace MODEL_ID in both places. Run openclaw models list to check registration. A listed model proves configuration discovery, not a successful API call.

​
Check the connection

Use an exact model ID available to your key, and begin with a short text request. For coding agents, first ask for a reply without reading or changing files or running commands. A successful text reply does not verify every tool or multimodal feature. For 401, check the key and its status. For 404, check the address and protocol. For model errors, recheck the model ID and selected group. See HTTP troubleshooting. Upstream documentation.