MCP Sora

What mcp-sora does

mcp-sora MCP server connects MCP clients to Sora video-generation actions through the AceDataCloud API. It provides tools for creating videos from text prompts, animating reference images, and generating videos that include a character from a reference video. The server also exposes task-status and discovery operations so a client can inspect running jobs and available API capabilities.

The project is listed in the 🎥 Multimedia Process category. It is relevant for agent workflows that need to submit video-generation requests or check the results of previously submitted tasks from an MCP-compatible application.

How it works

The hosted deployment uses the streamable HTTP endpoint https://sora.mcp.acedata.cloud/mcp. Requests use an AceDataCloud API token as a Bearer token. Claude.ai can connect through its OAuth flow, while other documented clients can include the token in their MCP configuration.

A local installation runs through the mcp-sora command. It can communicate over stdio for local desktop clients or start an HTTP listener with a selected port. The server sends generation requests to AceDataCloud, and clients can use the task tools to check individual or multiple jobs. Asynchronous generation tools accept a callback-oriented workflow.

Setup and configuration

For the local package, install mcp-sora from PyPI with pip, or run it through uvx. Set ACEDATACLOUD_API_TOKEN before starting the process. The default command runs the server in stdio mode; mcp-sora --transport http --port 8000 starts HTTP mode.

The README also documents a hosted option that does not require local installation. Configure the MCP URL and an Authorization: Bearer YOUR_API_TOKEN header in clients such as Claude Desktop, Cursor, Windsurf, Cline, and other listed MCP applications. Docker self-hosting is documented with the image ghcr.io/acedatacloud/mcp-sora:latest and port 8000.

Tools and capabilities

The mcp-sora MCP server exposes these operations:

• sora_generate_video creates a video from a text prompt.
• sora_generate_video_from_image creates a video using reference images.
• sora_generate_video_with_character uses a character from a reference video.
• sora_generate_video_async starts generation with callback notification.
• sora_generate_video_v2 and sora_generate_video_v2_async target Sora Version 2 through the partner channel.
• sora_get_task retrieves one task’s status and result.
• sora_get_tasks_batch checks several generation tasks at once.
• sora_list_models reports available Sora models and their capabilities.
• sora_list_actions reports available API actions and their corresponding tools.

The README also identifies landscape and portrait video orientations as supported features, although the supplied tool descriptions do not specify the parameter names used to select them.

Limitations and notes

The repository README marks the integration as retired and says it is no longer actively offered. That status should be considered before adopting it for a new production workflow. Video generation depends on access to the AceDataCloud API and a valid API token when using the documented token-based setup. The supplied material does not specify pricing, generation quotas, model-specific parameters, callback payloads, or result-retention behavior.

Getting started with this mcp-sora MCP server

Always refer to the official documentation for the most accurate and up-to-date information.