October 2, 2026 · 4 min read
YouTube Transcript MCP Server for Claude, Cursor, and Codex
AI coding agents cannot watch YouTube, and fetching a watch page returns no transcript. An MCP server gives the agent a tool it can call instead: paste a video link, and it pulls the transcript itself. TranscriptYT hosts one at transcript-yt.com/mcp that uses the same API key and credits as the REST API.
Add it to Claude Code
Create a free key in the dashboard, then run one command. Replace ts_live_... with your key.
claude mcp add --transport http transcriptyt https://transcript-yt.com/mcp \ --header "Authorization: Bearer ts_live_..."
Add it to Cursor, VS Code, or Codex
Clients that read a JSON or TOML config take the server URL and an Authorization header. Cursor uses .cursor/mcp.json; VS Code uses .vscode/mcp.json with type "http"; Codex uses ~/.codex/config.toml. The docs have the exact snippet for each, plus Claude Desktop via mcp-remote.
// .cursor/mcp.json
{
"mcpServers": {
"transcriptyt": {
"url": "https://transcript-yt.com/mcp",
"headers": { "Authorization": "Bearer ts_live_..." }
}
}
}What the agent gets
The server exposes three tools:
- get_transcript: a transcript by URL or ID, with optional language, translate_to, format (Markdown by default), and timestamps. 1 credit.
- list_languages: the caption tracks a video has. Free.
- get_usage: your remaining credits. Free.
Things to ask
Once connected, paste a link and ask in plain language. The agent decides when to call the tool.
- Summarize this video in five bullets.
- Turn this tutorial into a step-by-step checklist.
- Build the app shown in this video.
- List every product mentioned in this review, with timestamps.
MCP or the REST API?
Use the MCP server when a person is chatting with an agent and wants it to read videos on demand. Use the REST API when your own code needs transcripts in a pipeline, a backend, or a product feature. Both share the same key, credits, and video coverage, including AI transcription for videos without captions.