Social → Context

What Social → Context MCP server does

Social → Context accepts links from TikTok, Instagram reels, posts, and carousels, YouTube videos, and direct .mp4 or image URLs. It turns media into information an AI client can process, including author, caption, duration, engagement counts, audio or video transcripts, and visual descriptions. For posts containing multiple images, visual output can include OCR and a summary for each image.

The service is read-only. It does not modify the source post or account. You can use the Social → Context MCP server when an agent needs to summarize a reel, extract a caption and transcript, inspect text embedded in carousel images, or answer questions about a linked video.

How it works

The server is hosted at https://api.mcp.ai/p_socialdl and communicates over MCP HTTP. An MCP client sends a supported social link to one of the available tools, and the service returns structured content for the client’s model.

YouTube extraction uses existing captions when they are available. For other supported media, extraction can include audio or video transcription and image analysis. The metadata path avoids transcription and vision processing, while the download path resolves source CDN URLs instead of producing text.

Returned data is sent to the LLM host selected by the user, such as Claude, ChatGPT, Cursor, or another agent host. The documentation recommends considering the host’s training opt-out settings when handling content.

Setup and configuration

Add the hosted endpoint as a custom MCP connector. Claude Web and Claude Desktop can use the connector URL, while Cursor and VS Code provide installation links in the project documentation. Other clients can connect if they support MCP over HTTP.

No environment variables or local service configuration are documented for the hosted endpoint. The first three calls are free; subsequent usage is billed according to the service’s pricing documentation. The public repository is licensed under MIT, while the hosted MCP service itself is proprietary.

Tools and capabilities

• socialdl_extract processes a supported link into metadata, transcript, and visual output. It is the main option for giving an agent the contents of a video or post.
• socialdl_metadata performs a lighter lookup for metadata and media URLs without transcription or image analysis. Use it when the caption or basic post details are sufficient.
• socialdl_download returns direct source-CDN URLs for video qualities, audio tracks, and images. It is intended for saving media rather than interpreting it.

Together, these tools let the Social → Context MCP server separate inexpensive metadata lookups from transcript and vision processing, or from media retrieval.

Limitations and notes

The service only reads from the source and does not provide tools for editing or publishing content. Supported inputs are limited to TikTok, Instagram reel/post/carousel links, YouTube links, direct .mp4 files, and direct image URLs as described in the documentation. The repository is a public wrapper containing manifests, documentation, and skills; the hosted implementation is not open source.