Use Loralume inside Claude
Connect once, then ask for pictures of the people you've trained — in the same conversation where you're already working. No tab, no prompt-copying.
Ask Claude for something like:
“Make a photo of Michael and Raichelle on a beach at sunset, portrait orientation.”
Claude looks up your trained subjects, renders it on Loralume, and hands the image back in the conversation. The Sparks come out of your balance.
Add the connector
In Claude — Settings → Connectors → Add custom connector
Paste the Loralume server URL. Claude will send you here to sign in and approve access.
https://loralume.com/mcpIn Claude Code — one command
Same connector, from the terminal.
claude mcp add --transport http loralume https://loralume.com/mcpApprove, and you're connected
You'll sign in with your Loralume account and see exactly what you're granting. You can disconnect any time from the same screen in Claude.
What Claude can do once connected
The people, pets and characters you've trained — by name, not by ID.
Single subjects or a whole Cast in one scene, then refine in conversation.
Claude can tell you what a render cost and what's left.
Training a new subject still happens here on the site — it needs photos uploaded from a browser. Claude will hand you the link when you ask.
What you need
- · A Loralume plan — the connector is included on Spark+, Pro and Studio.
- · At least one trained subject, if you want images of a specific person.
- · Sparks in your balance. Renders cost the same here as they do on the site.
Loralume speaks the Model Context Protocol, so the same connection works in any MCP-capable client — not just Claude.