Plain-language image and video generation for people and agents. REST API and hosted MCP server.

Maginary is an image and video generator you talk to in plain language, in any language. There's no prompt syntax to learn: say what you want and you get four takes back in a few seconds. It routes between ~20 model families (Flux, Nano Banana, GPT Image, Ideogram, Recraft, Kling, Seedance, Sora and others) so you don't have to pick one. For agents there's a hosted MCP server at mcp.maginary.ai/mcp that works with Claude, Cursor and other MCP clients, plus a REST API with webhooks. Agents can also pay per call over x402 without an account. You can edit an image by pasting its URL with an instruction, upscale it, or turn it into a short video.