Luw.ai

What Luw.ai MCP server does

The Luw.ai MCP server exposes Luw.ai’s AI design and media workflows to MCP-compatible clients. It is intended for visual tasks such as redesigning interiors, staging empty rooms, rendering architectural sketches, editing images, creating product scenes, generating videos, and producing 3D models from reference photos.

The service can work from image URLs through its hosted endpoint or from local file paths when the local server is used. Example requests include redesigning a bedroom in a chosen style, changing a kitchen floor material, removing furniture, turning a render into a drone fly-through video, and creating a seamless tile texture.

How it works

The hosted endpoint is available at https://mcp.luw.ai/mcp and uses Streamable HTTP. Authentication can use MCP OAuth through client sign-in or a bearer token containing a Luw.ai API key. The hosted server is stateless and passes the key to the Luw.ai API for each request; it does not read files from the local computer.

For local file access, the npm package runs through npx on the user’s machine. The local process receives LUW_API_KEY and can work with paths such as ~/Desktop/room.jpg. This option is appropriate when prompts need to reference files that are not available at public URLs.

The Luw.ai MCP server also provides built-in prompts that appear as slash commands or prompt templates in supported clients, including redesign_room, stage_empty_room, sketch_to_render, and product_shot.

Setup and configuration

The simplest setup is to add https://mcp.luw.ai/mcp to a client with remote MCP support, then sign in to Luw.ai. New accounts receive free credits. Clients that cannot use sign-in can send Authorization: Bearer YOUR_LUW_API_KEY instead. Passing the API key in the URL is also documented, but the URL must remain private because anyone who obtains it can spend the associated credits.

For local use, install Node.js 18 or newer and provide an API key from the Luw.ai API page. The package can be launched with npx -y @luw-ai/mcp. Claude Desktop also supports the downloadable luw.mcpb package, which does not require a separate Node.js installation.

The Luw.ai MCP server has documented setup paths for Claude, Claude Code, Claude Desktop, Cursor, VS Code with Copilot, Windsurf, OpenAI Codex CLI, Gemini CLI, and Kiro. Some clients can keep the key in secure or secret storage.

Tools and capabilities

• Redesign rooms in a selected style and generate multiple variations.
• Virtually stage empty rooms.
• Render sketches as photorealistic scenes.
• Change image elements such as flooring or furniture, including floor segmentation for material replacement.
• Remove furniture and create product-shot scenes.
• Generate fly-through or other video output from renders.
• Create 3D models from multiple reference photos.
• Generate seamless material textures at a requested size.
• Use luw_interior_design for room redesign and empty-room staging.

Limitations and notes

The hosted server cannot read files on the user’s computer, so inputs must be supplied as image URLs or processed through the local server. Local operation requires Node.js 18 or newer and a Luw.ai API key.

Usage consumes Luw.ai credits. The README states that a new account includes free credits, while the documented interior-design tool uses one credit. The available credits, API key, and disconnection controls are managed through the Luw.ai account and API page.

Remote access requires a client that supports Streamable HTTP, MCP OAuth, bearer headers, or the documented API-key URL fallback. Keep API-key URLs private because they authorize requests against the associated account.