Youtube Transcripts

What youtube-transcripts MCP server does

The youtube-transcripts MCP server exposes the Apify johnvc/YoutubeTranscripts Actor through an MCP endpoint. It extracts captions from one or more YouTube URLs and returns one structured record per video. The output includes the video ID and URL, transcript language and language code, whether captions were auto-generated, total duration in seconds, and a success indicator.

Each successful record contains two transcript representations. timestamped is an array of caption segments with text, start time, and duration. non_timestamped contains the complete transcript as one plain-text string. This supports both time-aware applications, such as synchronized interfaces, and text-oriented workflows, such as search or analysis.

How it works

Provide youtube_url as either one URL string or an array of URLs. The Actor supports standard YouTube video links, Shorts, youtu.be links, embed URLs, and mobile URLs. Multiple inputs are processed in parallel. If one URL is invalid or unavailable, its result records an error and success: false rather than stopping the complete run.

The MCP connection is hosted at https://mcp.apify.com/ with query parameters that preload the Actor and the related actor and documentation tools. Authentication can use Apify OAuth in a browser or an Apify API token.

Setup and configuration

The youtube-transcripts MCP server requires an Apify account and API key. For the repository’s Python example, the documented prerequisites also include Python 3.11 or newer and the UV package manager. The token can be stored in a .env file based on .env.example, or exported as APIFY_API_TOKEN before running the example.

For remote MCP use, configure this endpoint:

https://mcp.apify.com/?tools=actors,docs,johnvc/YoutubeTranscripts

Claude Desktop can load it through mcp-remote, while Claude Code can register it with the HTTP transport. Cursor uses a project-level .cursor/mcp.json configuration. Authentication is completed through OAuth when prompted, or by supplying the Apify token.

Tools and capabilities

• Accept one YouTube URL or a batch of URLs.
• Process batch inputs in parallel.
• Return caption text with per-segment timing.
• Return a single plain-text transcript.
• Report language, language code, and automatic-caption status.
• Include video duration and per-video success or error status.
• Handle standard, Shorts, youtu.be, embed, and mobile URL formats.

The MCP endpoint is scoped to the johnvc/YoutubeTranscripts Actor alongside the actors and docs tools shown in the setup URL. The underlying repository also includes a Python invocation example using UV.

Limitations and notes

A transcript must be available for a successful result; unavailable or invalid URLs produce an error note and success: false. The service is dependent on Apify authentication and usage billing.

The youtube-transcripts MCP server returns the transcript data generated by the Actor. Caption text is represented as a placeholder in the repository’s example output, while an actual run returns the retrieved caption content.