> For the complete documentation index, see [llms.txt](https://docs.roboflow.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.roboflow.com/changelog/explore-by-month/july-2026/mcp-server-upload-custom-model-weights-from-your-ai-assistant.md).

# MCP Server: Upload custom model weights from your AI assistant

You can now upload locally trained model weights to Roboflow directly from Claude Code, Cursor, or any MCP client, with no manual packaging or upload scripts. Supported architectures include YOLOv5 through YOLOv12, YOLO26, YOLO-NAS, RF-DETR, Florence-2, and PaliGemma.

Weights can be uploaded as a versioned deployment tied to a dataset version, or registered as a workspace model. Incompatible architectures, missing files, and invalid uploads return clear validation errors instead of failing silently.