
"Create a cinematic image of a crowded Thai night market with a neon Pad Thai sign."
Create images, videos, and audio directly from your prompts in your AI agent.
https://mcp.magichour.ai/In Claude desktop or claude.ai, go to Settings → Connectors, add a custom connector named Magic Hour, paste the URL, and set the OAuth Client ID to magic-hour-mcp.
Open Claude connectorsConnect, paste your Magic Hour API key, then ask Claude to generate an image, video, or audio.
Start creating in ClaudeThe MCP server exposes Magic Hour's current API tools. It also tells the agent how to upload files, wait for renders, and return download links.
Generate, edit, colorize, remove backgrounds, and swap faces or heads.
Create from text or stills, lip sync, animate portraits, and swap faces.
Generate speech. Use audio files with talking-photo and other tools that accept audio.

"Create a cinematic image of a crowded Thai night market with a neon Pad Thai sign."
"Animate this illustrated hooded character with slow camera movement."

"Add stylish sunglasses to this photo and keep the result realistic."
The first call returns a project ID. A Magic Hour wait tool checks the job and returns the exact signed download URL when the file is ready.
Agent calls the matching Magic Hour tool.
Project polls until a terminal state.
Agent uploads source files before the render.
Exact download URLs come back.
These answers match the current MCP repo. We leave out anything the code does not confirm.
https://mcp.magichour.ai/