Sora

The Sora MCP Server integrates with OpenAI’s latest Sora 2 API to help you generate and remix videos directly from an MCP-compatible client.

You can feed it text prompts to create new video clips, modify existing videos, and keep tabs on the generation progress without leaving your workflow.

Features

• 🎥 Create Videos: Generate video clips directly from text descriptions using Sora 2.
• 🔀 Remix Videos: Take an existing video and alter it with a new prompt to create variations.
• 📊 Track Job Status: Check the real-time progress of your video generation tasks.
• 💾 Auto-Download Videos: A handy tool automatically saves completed videos to your local machine.
• 🔌 Dual Server Architecture: Includes separate implementations for local (stdio) and remote (http) connections, optimizing for security and accessibility.

How To Use It

1. Clone the repository and install the necessary dependencies:

git clone https://github.com/Doriandarko/sora-mcp
cd sora-mcp
npm install

2. Build the project.

npm run build

3. This project contains two different MCP server implementations for distinct situations:

• stdio-server.ts: Designed for local clients like Claude Desktop. It communicates via standard input/output (stdio), which is fast, secure, and doesn’t require a network connection.
• server.ts: Built for remote access from web-based tools or clients like VS Code and Cursor. It runs as an HTTP server on port 3000.

4. Connecting to your MCP clients

For Claude Desktop:
You need to create a configuration file to tell the Claude app how to run the server.

1. Copy the example config file claude_desktop_config.example.json to ~/Library/Application Support/Claude/claude_desktop_config.json.
2. Open the new file and update the args path to the absolute path of the sora-mcp/dist/stdio-server.js file on your machine.
3. Add your OpenAI API key to the OPENAI_API_KEY field.
4. Restart the Claude Desktop application. The Sora tools will now be available.

For HTTP Clients (VS Code, MCP Inspector, etc.):
Run the server in development mode.

npm run dev

Or run it in production mode.

npm run build
npm start

You can then connect your client to http://localhost:3000/mcp.

5. All available tools

• create-video: Generates a video from a prompt. You can specify the model, duration, and resolution.
• get-video-status: Checks the progress of a video generation job using its video_id.
• list-videos: Retrieves a paginated list of all your video generation jobs.
• download-video: Returns a curl command to download a completed video file manually.
• save-video: Automatically downloads and saves a completed video to your computer. This is often more convenient than using download-video.
• remix-video: Creates a new video based on an existing one with a new text prompt.
• delete-video: Removes a video job and its associated files.

FAQs

Q: Do I need to manually download the video after it’s generated?
A: You can, using the download-video tool that gives you a curl command. However, a more direct workflow is to use get-video-status to confirm the video is complete, and then use the save-video tool to automatically download it to your specified directory.

Q: How do I update the server path in the Claude Desktop configuration if I move the project folder?
A: You must edit the claude_desktop_config.json file located in Claude’s Application Support directory. Update the absolute path in the args field to point to the new location of dist/stdio-server.js and restart Claude.

Q: What do I need to start using the Sora MCP Server?
A: You’ll need Node.js 18+, an OpenAI API key with Sora API access, and an MCP client like Claude Desktop or Cursor. The Sora API requires separate approval from OpenAI beyond a standard API account.

Q: How much does it cost to generate videos?
A: You pay standard OpenAI API rates for Sora usage based on video duration and resolution. The MCP server itself is open-source and free, with pricing determined by your OpenAI API consumption.