OpenClaw-ready
OpenClaw uses the same Luma Cloud API contract.
Set the public /v1 endpoint, a workspace API key, and a canonical model id. Keep private gateway, bridge, shard, and local OmniRoute details out of client-facing setup.
Open API setup
Use the public Luma Cloud /v1 endpoint from App -> API. Local dev commands are only for internal smoke work, not for customer provider setup.
Verify the branded endpoint
Call `https://api.lumaos.cloud/healthz` first, then sign in to `/app` so the current session gets a workspace-scoped starter key.
Copy the workspace key
Open the customer app, sign in, and copy the included starter key from Home or Access. Do not rely on the shared local builder seed for onboarding.
Configure an OpenAI-compatible client
Set `baseURL=https://api.lumaos.cloud/v1`, use the workspace key from `/app`, and start with model `gpt-5.5-medium` before switching to `gpt-5.5-xhigh`.
Confirm request-id support
Keep the x-lumacloud-request-id or error.request_id from the first failed request. Support works from public ids, not from private bridge details.
Provider setup
Use the branded base URL exactly as the provider surface.
The client should only know Luma Cloud. OmniRoute, bridge hosts, and private topology remain hidden behind the gateway boundary.
Connection values
Local alpha defaults for the first OpenClaw-style provider hookup.
Base URL
https://api.lumaos.cloud/v1
Access key
Sign in to `/app` and copy the included starter key for the current workspace.
Health check
https://api.lumaos.cloud/healthz
Local alpha truth
What is already real versus what stays future work.
Already supported
Still intentionally missing