Support

ImageLoRA answers questions about training image-model LoRAs inside your AI assistant. It covers FLUX, Krea 2, Z-Image, SDXL, Qwen-Image, MOSS Image, Fibo 2.0 and more, using the same data as imagelora.dev.

Connect it

Server URL: https://imagelora.dev/mcp (Streamable HTTP, no sign-in).

ChatGPT: find ImageLoRA in the apps directory and connect it.
Codex: codex mcp add imagelora --url https://imagelora.dev/mcp
Cursor, Claude and other MCP clients: add a remote MCP server with the URL above.

What you can ask

Which trainers can train a LoRA for a given model.
Training settings for a model, optionally for one trainer.
Whether your VRAM is enough for a model.
How two trainers compare, such as ai-toolkit and OneTrainer.
The cheapest cloud GPUs with a given amount of VRAM.
Recent news on LoRA training for a model or trainer.

Every answer includes a link to the matching imagelora.dev page and the original sources.

If something doesn't work

The assistant doesn't use ImageLoRA: name it in your message, for example "Use ImageLoRA to check if 12 GB is enough for a FLUX.2 klein LoRA."
"No data for this model": ImageLoRA covers image models only. See the list on the home page.
Prices look old: each price answer shows when prices were last fetched. They refresh about every 30 minutes.

Report wrong or outdated data

Email [email protected] with the model, the trainer and a link to the source that shows the correct information.

Contact

[email protected]. We aim to reply within a few days.

Last updated 25 October 2026. · Privacy · Terms · Support