MCP server for AI image, video and audio generation | Visual Sandbox
Claude Code
$
get started — install the vsb cli
curl -fsSL visualsandbox.com/install.sh | sh && vsb setup

Or paste the prompt into Claude Code — it installs the CLI, signs you in, and runs your first render.

Using an AI agent? Copy a setup prompt and it connects itself.

CLIMCPClaude Plugin
MCP server

The Visual Sandbox MCP server

One endpoint. Anything that speaks MCP can connect.

29 modelsNo API keyPay per run

It already knows how to prompt them

Every model ships with a prompting guide. Your agent reads it before it writes.

How to set up

Set it up in
Claude CodeCodexCursor
claude mcp add --transport http visual-sandbox https://visualsandbox.com/mcp
https://visualsandbox.com/mcp

The browser signs you in. No key to paste.

Paste this into Claude Code
Connect to the Visual Sandbox MCP server (visualsandbox.com) so you can generate images, video and audio for me.

1. Add the server: claude mcp add --transport http visual-sandbox https://visualsandbox.com/mcp
2. Run /mcp, pick visual-sandbox, and open the sign-in. Do not set an Authorization header — it registers your client itself. Ask me before you approve anything on my behalf.
3. Run /mcp again and confirm visual-sandbox is connected. If it is not, stop and tell me what failed.
4. Do not generate anything yet. Every run costs credit, so ask me what to make, price it with estimate_cost, and wait for my yes.

Frequently asked

There is no subscription. You pay for each run — the provider cost plus a small platform markup — from a balance you top up.