Summary
One connection to a lot of generation models, priced in a currency you can check before you spend it.
Because generation is asynchronous and costs credits, the two supporting calls matter as much as the generators: list_models tells you what a given model costs right now, and get_balance tells you what is left — so an agent can pick a model on price rather than on habit.
What it is
A hosted generation server: text or image in, and video, image, audio or a GLB 3D model out, with job polling and inline rendering of the result.
What you get
- Text-to-video and image-to-video, with Veo, Seedance, Kling, Wan and Sora among the models
- Text-to-image and image editing across Flux, Imagen, Ideogram and GPT Image
- Speech, music, sound effects and voice cloning through ElevenLabs, MiniMax Speech HD, Gemini Flash TTS, Stable Audio and Mirelo SFX
- Text-to-3D and image-to-3D, returned as GLB
- Asynchronous jobs: each generate call returns a task ID you poll, then render inline
- Your past generations listed, your credit balance readable, and the live per-model credit cost queryable
Requirements
An Arcframe account, authorized over OAuth — there is no key to copy. It connects over Streamable HTTP at https://mcp.arcframe.ai/mcp. Generation spends credits; the free plan covers image generation, while video, audio and 3D need a paid plan.
Setup effort
One command — npx -y mcp-remote https://mcp.arcframe.ai/mcp