Skip to main content
Glama

Stivio MCP

Turn a photo into a short AI video from inside your AI client. This repo holds setup notes for the hosted Stivio MCP server: list video models, get an exact credit quote, upload an image, start a render and fetch the MP4.

  • Endpoint: https://stivio.ai/api/mcp (Streamable HTTP)

  • Auth: Authorization: Bearer <key>, key from stivio.ai/mcp

  • Official MCP Registry: io.github.wsmhj/stivio-mcp

The server is hosted, so most clients connect to the URL directly. For clients that only run local commands there is a small stdio bridge in this repo (npx -y github:wsmhj/stivio-mcp, source in bin/). The server code itself is not in this repo.

Tools

Tool

What it does

Key needed

list_video_models

Models, supported settings and per-second credit rates

No

estimate_video_cost

Checks settings and returns the exact credit cost

No

get_credit_balance

Your spendable credit balance

Yes

upload_image

Uploads a PNG, JPEG or WebP (max 3 MiB) and returns a URL to animate

Yes

generate_video

Starts a paid render from an image or a text prompt

Yes

get_video_status

Polls your job; returns the MP4 URL when done

Yes

Model listing and quotes work without a key, so you can try the server before signing up. Generation spends credits from your Stivio account, and failed renders are refunded automatically.

Related MCP server: Agnes Video MCP Server

Get a key

  1. Sign in at stivio.ai/mcp.

  2. Create a personal key. Keys last 30 days and can be revoked on the same page.

  3. Paste it into your client config below in place of YOUR_STIVIO_KEY.

Setup

Claude Code

claude mcp add --transport http stivio https://stivio.ai/api/mcp \
  --header "Authorization: Bearer YOUR_STIVIO_KEY"

Cursor

~/.cursor/mcp.json (or .cursor/mcp.json in a project), also in examples/cursor.json:

{
  "mcpServers": {
    "stivio": {
      "url": "https://stivio.ai/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_STIVIO_KEY" }
    }
  }
}

VS Code

.vscode/mcp.json; VS Code asks for the key once and stores it. See examples/vscode-mcp.json.

Claude Desktop and other stdio-only clients

Clients that only launch local commands can run the small bridge in this repo; it forwards everything to the hosted server (Node.js 18+):

{
  "mcpServers": {
    "stivio": {
      "command": "npx",
      "args": ["-y", "github:wsmhj/stivio-mcp"],
      "env": { "STIVIO_API_KEY": "YOUR_STIVIO_KEY" }
    }
  }
}

Without STIVIO_API_KEY the bridge still lists models and gives quotes.

Other clients

Any client that supports remote Streamable HTTP servers with a custom header works. Clients that only accept OAuth connectors are not supported yet.

Example prompts

  • "Which Stivio models can do 10-second videos, and what would 10 seconds at 720p cost?"

  • "Upload ~/Pictures/beach.jpg and animate it: slow push-in, waves rolling in, 5 seconds. Show me the quote first."

  • "Check my Stivio video status and give me the MP4 link."

The server tells the assistant to show you the prompt, settings and credit cost and wait for your OK before it calls generate_video.

Uploading local files from a shell

Clients with shell access can upload a file directly and pass the returned url to generate_video:

curl https://stivio.ai/api/upload \
  -H "Authorization: Bearer YOUR_STIVIO_KEY" \
  -F "file=@photo.jpg"

The docs and examples in this repo are MIT licensed.

Related MCP Connectors

Related MCP Servers