Make Seedance 2.5 videos from Claude, ChatGPT or Cursor

Connect it

Any other MCP client that takes a URL and headers works with an API key from your dashboard:

{
  "mcpServers": {
    "revana": {
      "type": "http",
      "url": "https://revana.si/api/mcp",
      "headers": {
        "Authorization": "Bearer sr_live_YOUR_KEY"
      }
    }
  }
}

Then just ask

Make a 10 second 720p Seedance 2.5 video of a paper boat sailing down a rainy street, with the sound of rain. Tell me the price first.

The assistant calls get_quote and tells you the price ($2.10 for that clip), waits for your yes, calls generate_video with that price as the cap, follows it with get_video, and gives you a link from get_video_link. To animate a picture or use references, attach it: upload_file turns it into a link the model can use.

The tools

Spending stays in your hands

• Videos are paid from credit you top up in advance; there is no bill afterwards.
• Every video carries a cap (max_cost_usd, up to $100 a video): it never costs more.
• The assistant is told to quote and ask before it spends. Retrying a request with the same id never makes or charges a second video.
• Failed videos are not charged.
• Revoke a connection any time by revoking its key in your dashboard.

Why MCP rather than a skill or a script

A skill file teaches an assistant to call an API, but it still needs a key, a place to run code and somewhere to keep the file. The MCP server is hosted: the assistant gets typed tools, sign-in is a button, and nothing runs on your machine. It is listed in the official MCP registry as si.revana/revana. For code you write yourself, the REST API takes the same requests.