Stack
CLI, MCP, n8n, Dify, WordPress, Adobe, Blender, ComfyUI, Python, JavaScript, Vercel AI SDK. Same key as the studios.
Every generation is async. POST /api/v1/{endpoint} with x-api-key, receive a request_id, poll GET /api/v1/predictions/{id}/result until completed. Optional ?webhook= on submit. Costs come back on the payload. Sandbox keys (is_test) return mocks.
The 20+ apps share Next.js, Prisma, Google OAuth, Stripe credits, and webhook-backed MuAPI. Env: DATABASE_URL, NEXTAUTH_*, GOOGLE_*, MUAPIAPP_API_KEY, STRIPE_*, WEBHOOK_URL. Fork, brand, deploy to Vercel.
White-label is hosted by MuAPI ($49/$149/$499). You never run a server. Forks are yours: you run Next, Stripe, and Postgres. Use both — white-label for speed, a fork when you want the code.
Hosted, branded MuAPI under your domain.
One webhook across Fal, Replicate, MuAPI.
Your Comfy-style engine. Self-owned, air-gapped, audited.
Your ComfyUI, on MuAPI.
600+ models as Comfy nodes.
Seedance 2.0 / 2 Mini nodes.
Veo 3.1 inside the graph.
Generate, remix, extend music.
V7, V8, Niji as nodes.
Any-to-any video nodes.
GPT Image in the graph.
Alibaba’s 1080p video model.
Opus Clip, as nodes.
Claude, Cursor, any MCP client.
Drop-in skills for Claude Code, Cursor, Gemini CLI.
Typed Python client.
Browser and Node client.
MuAPI inside the AI SDK.
Seedance 2.0 / Mini for n8n.
MuAPI inside Dify apps.
Generate media from Claude Code.
Generate inside wp-admin.
Photoshop and Premiere.
MuAPI on the 3D timeline.