DeepSRT

DeepSRT MCP Server provides YouTube video summarization and transcript extraction by integrating with DeepSRT’s processing flow and YouTube captions. It exposes simple tools you can call from an MCP client to generate summaries in multiple languages or extract timestamped transcripts, enabling fast, self-contained video analysis inside MCP-enabled environments.

How to use

You use the DeepSRT MCP Server by connecting an MCP client to the local stdio server. The server exposes two tools: get_summary for generating a formatted video summary and get_transcript for extracting a transcript with timestamps. Start the MCP server in stdio mode, then invoke the tools from your MCP client to process any YouTube video that has captions.

How to install

Prerequisites you need before starting: a runtime capable of running MCP servers and tooling to launch the server. The recommended approach is to run the MCP server in stdio mode using Bun as the runtime.

# Start the MCP server in stdio mode (recommended)
bunx @deepsrt/deepsrt-mcp@latest --server

Configuration and usage notes

The MCP server is designed for seamless integration with MCP clients such as Claude Desktop or Cline. The server runs locally and exposes two tools you can call via MCP: get_summary and get_transcript. For long-term usage, you can configure your MCP client to launch the server automatically using the provided stdio configuration.

{
  "mcpServers": {
    "deepsrt": {
      "type": "stdio",
      "command": "bunx",
      "args": [
        "@deepsrt/deepsrt-mcp@latest",
        "--server"
      ]
    }
  }
}

Examples of typical usage

From an MCP client, request a video summary or transcript by calling the corresponding tool and passing a video identifier. You can specify language preferences and mode for summaries. The server handles YouTube caption retrieval, selects the best caption track (manual captions preferred), and routes the transcript to DeepSRT for summarization when generating a summary.

Security and reliability tips

Ensure the server runs in a secure environment with network access limited to trusted IPA (internal) endpoints. The server uses timeouts and validation to provide friendly error messages and gracefully handle videos without captions. When running in production, monitor for rate limiting from YouTube and DeepSRT and apply suitable backoff strategies.

Notes on deployment and maintenance

Maintenance is straightforward: keep the MCP server up to date by using the latest release tag when starting via bunx. Restart the MCP server to pick up updates. No pre-caching is required; the server fetches video details and captions on demand.

Available tools

get_summary

Generates a formatted video summary for a YouTube video. Supports language and mode options to tailor the output.

get_transcript

Retrieves a timestamped transcript from YouTube captions for a given video. Returns transcripts with precise timestamps.