Scumble launches as open-source desktop editor for AI inpainting with MCP support

Scumble, a free and open-source desktop editor designed specifically for AI inpainting, has launched on Product Hunt. The tool streamlines the workflow of masking, generating, and integrating AI results by returning every generation as a separate, color-matched layer that preserves the original image.

Core Features and Workflow Users can select areas for inpainting by painting, pointing at objects, or typing descriptions. The editor supports:

• Layer-based results: Generated content lands as its own layer, color-matched to surroundings.
• Privacy-focused API usage: Only the area around the selection is sent to the API, with full-resolution results returned.
• Advanced prompting: Supports "boxes in the prompt" for spatial control (FLUX 3 Image, Ideogram 4) and reference images by name.
• Export options: Includes PSD, ORA, and TIFF export, along with 46 film looks and retouch tools.

Model and Infrastructure Support Scumble operates without its own backend, relying on user-provided infrastructure. It connects to:

• Local/Remote ComfyUI: Supports local instances, remote servers, or Comfy Cloud.
• API Providers: Compatible with FLUX 3, FLUX.2, GPT Image, Nano Banana, Seedream, Qwen, Ideogram, and others using user API keys.
• MCP Integration: Includes an MCP server with 105 tools, allowing AI agents like Claude Code to drive the editor programmatically.

Availability and Status The software is licensed under GPL-3.0 with no telemetry or account requirements. Version 0.1.x is currently available for Windows via the Microsoft Store, GitHub installer, and portable zip. Linux builds are generated via CI but remain untested by the developer, while macOS support is planned. The project explicitly states that Scumble takes no cut of API costs, as users pay providers directly.