VideoGen MCP

What VideoGen MCP does

VideoGen MCP MCP server connects MCP clients to the live VideoGen API rather than only exposing documentation. It provides tools for end-to-end video workflows, media generation, project management, file handling, entities, and account information. Supported workflows include turning scripts, voiceovers, slideshows, or storyboards into videos.

The media surface includes image, video clip, speech, sound-effect, music, and avatar generation. It also includes image and video background removal, image vectorization, image and video upscaling, and a 3D image effect. Project tools can list and retrieve projects, export them, remix them, and inspect remix actions.

This fits agents that need to create product demos, ecommerce advertisements, marketing videos, or other media projects through VideoGen. The server also supports uploading source assets and using ACTOR, PRODUCT, or VISUAL_STYLE entities in workflows.

How it works

VideoGen MCP MCP server wraps the official VideoGen SDK and sends authenticated requests to the VideoGen API. It has two transports with the same tool surface: local stdio and remote Streamable HTTP. In local mode, the MCP client launches the package as a subprocess. In remote mode, the hosted endpoint creates a request-specific server and receives credentials in the Authorization bearer header.

Remote long-running workflows, media operations, and exports return a run or execution identifier immediately. Agents can then call the relevant get tool to poll for completion. The local stdio mode waits for completion by default.

File handling varies by transport. Local clients provide a file path to upload. Remote clients can send small files as base64 data, use a short-lived pre-signed upload URL for larger files, or use the ChatGPT-only open_uploader widget. The server does not fetch caller-provided URLs.

Setup and configuration

For remote use, connect an MCP client that supports Streamable HTTP to https://mcp.videogen.io/mcp and provide a VideoGen API key or OAuth credential. This avoids installing the package. The standard endpoint requires credentials for requests; the ChatGPT-specific /mcp/chatgpt surface has separate discovery and billing-message behavior.

For local use, run npx -y @videogen/mcp from the client configuration and set VIDEOGEN_API_KEY. Keys are forwarded to the VideoGen API and are not persisted by the server. VIDEOGEN_BASE_URL can override the upstream API base URL. Remote deployments can additionally configure OAuth issuer settings and the OpenAI Apps challenge token.

Tools and capabilities

The VideoGen MCP MCP server exposes 50 hosted tools according to the supplied schema snapshot. Notable groups include:

• Workflows: script_to_video, storyboard_to_video, voiceover_to_video, and workflow run controls.
• Media: generate_image, generate_video_clip, text_to_speech, generate_music, and generate_avatar.
• Projects: listing, retrieval, export, remix, and export-status tools.
• Files: inline upload, large-file upload setup, file retrieval, listing, and ChatGPT’s open_uploader.
• Entities: create, update, archive, and attach references to actors, products, and visual styles.

Use tool discovery on the connected server for the latest definitions; the repository’s schema snapshot may change.

Limitations and notes

The hosted documentation MCP at https://docs.videogen.io/_mcp/server is a separate service that reads API documentation and does not execute VideoGen operations. open_uploader is useful only in clients that render MCP Apps UI; other clients should use the regular upload tools. Remote requests are stateless, so agents must use returned run or execution identifiers when following asynchronous work.