YouTube

What the YouTube MCP server does

The YouTube MCP server connects MCP-compatible AI clients to a hosted YouTube service. It provides 22 tools for finding and inspecting YouTube content across videos, channels, playlists, comments, feeds, hashtags, and community surfaces. The repository describes the service as supporting both reading and writing, although the listed tools are primarily retrieval-oriented and include channel video and Shorts POST variants.

Search can cover videos, channels, and playlists. Other tools retrieve a video's metadata, comments, related videos, playlist items, home feed, trending feed, Hype feed, or hashtag posts. Channel-specific tools expose home, about, videos, Shorts, playlists, community, featured channels, store, and channel search data.

How it works

The hosted endpoint is available at https://api.mcp.ai/p_youtube. An MCP client connects to that HTTP endpoint and presents the available tools to an agent.

Requests can be made through the individual tools or through youtube_api, described as an escape hatch for calling any path on the YouTube backend. Several channel tools support bulk execution by accepting IDs, while youtube_resolve_url converts a YouTube URL or share link into a video ID and page type. youtube_download returns stream or download metadata, including formats, audio and video URLs, and status.

Setup and configuration

Connect the server using the hosted URL rather than running a local process. Claude Web and Claude Desktop can add it through Claude's custom connector settings. Cursor and VS Code have installation links in the README, and other MCP clients can use the same HTTP endpoint if they support MCP over HTTP.

No environment variables are listed as required for the hosted server. The repository is an MIT-licensed wrapper containing manifests, documentation, and skills; the hosted MCP service itself is proprietary.

Tools and capabilities

The YouTube MCP server includes tools for:

• Searching videos, channels, and playlists with youtube_search.
• Inspecting video details, comments, related videos, and resolved URLs.
• Reading home, trending, Hype, hashtag, and playlist feeds.
• Reading channel metadata, videos, Shorts, playlists, community posts, featured channels, and store data.
• Searching within a channel's videos and playlists.
• Retrieving stream and download metadata for a video.
• Calling a backend path through youtube_api.

Limitations and notes

The service is hosted rather than fully open source. The public repository is a wrapper, while the endpoint is proprietary. Data returned by the tools is sent to the selected LLM host, such as Claude, ChatGPT, Cursor, or a custom agent, and the README identifies that host as a subprocess outside the service's control. Review the provider's training opt-out options before sending sensitive requests or content.

Pricing starts with a free tier, with further plan details referenced in the repository documentation. Use this server when HTTP-based MCP access is acceptable and the agent needs YouTube-specific retrieval or channel exploration. It may be unsuitable where local-only processing or direct control over the backend is required.