Replace YOUR_KEY with your free API key. Create one here: sign in with Google, 10 requests a day free.

Or view the instructions as plain text.

Research a topic, not just one video

Reading one transcript answers one video. These tools let your AI work through dozens and report back with sources.

• search_youtube_videos finds videos on a topic, with channel, views and length. Filter out the short and the obscure.
• find_in_transcript returns only the passages that mention your keywords, each with a timestamp and a link to that moment.
• get_transcripts_batch reads up to 5 videos in one call, each cut to the length you choose.
• get_video_comments reads the top comments, where viewers say what worked for them.
• list_channel_videos lists the videos of a channel or playlist, 50 at a time, and pages on to the first 500.
• get_account_status shows your plan, daily quota and how many requests you have left. It is free and uses no quota, so your AI can check before a big job.

And a library of what people say, with full-text search: about 100 million spoken sentences from 425,000+ English videos, and typically 1 to 8 million new sentences added every day:

• search_spoken_words finds the sentences where people say something, with a link to that moment in the video.
• get_spoken_context reads the lines before and after a moment.
• rank_videos_by_topic and rank_channels_by_topic show which videos and channels talk about a topic the most.
• spoken_mentions_over_time counts mentions by day, week or month, to see whether a topic is rising.

Try asking:

Research how people improved their English speaking. Search YouTube in English and Chinese, look at videos from people telling their own story, find what methods they used and how long it took, check the comments, and tell me which methods several unrelated people agree on.

A research run like this takes about 30 to 40 requests: a few searches, then a look inside each promising video. The free key's 10 a day covers a quick look. Pro gives you 1,000 a day for $120 a year, enough to run several full research sessions a day without counting.

See it in action

Get your API key

The server needs a key to fetch transcripts. Create a free key (Google sign-in, about a minute) and put it in the connector URL below in place of YOUR_KEY. Already connected without a key? Remove the connector and add it again with the new URL.

Set it up in Claude

1. Open Settings → Connectors and click Add custom connector.
2. Set Name to youtube-transcript.ai and the URL to https://youtube-transcript.ai/mcp?key=YOUR_KEY, then save.
3. Start a chat and say, e.g. “Summarize this video: <YouTube link>”. Claude calls the get_youtube_transcript tool automatically.

Set it up in ChatGPT

1. In a Developer/Custom connector setup, choose Add MCP server.
2. Set Name to youtube-transcript.ai and the server URL to https://youtube-transcript.ai/mcp?key=YOUR_KEY (transport: Streamable HTTP).
3. Enable it for your chat, then paste a YouTube link and ask for a summary or transcript.

Set it up in Meta Muse

1. In a Muse chat, paste https://youtube-transcript.ai/mcp and ask Muse to connect this MCP server. It sets up the connection for you.
2. When Muse asks for your API key, give it the key you created above.
3. Paste a YouTube link and ask for a summary or transcript.

Set it up in Claude Code

Run this once in a terminal. --scope user makes the server available in every project.

claude mcp add --transport http --scope user youtube-transcript \
  https://youtube-transcript.ai/mcp \
  --header "Authorization: Bearer YOUR_KEY"

Set it up in Cursor

Add the server to ~/.cursor/mcp.json, or to .cursor/mcp.json inside one project.

{
  "mcpServers": {
    "youtube-transcript": {
      "url": "https://youtube-transcript.ai/mcp",
      "headers": { "Authorization": "Bearer YOUR_KEY" }
    }
  }
}

Set it up in Codex

Add the server to ~/.codex/config.toml.

[mcp_servers.youtube-transcript]
url = "https://youtube-transcript.ai/mcp"
http_headers = { "Authorization" = "Bearer YOUR_KEY" }

Any other MCP client

Give it https://youtube-transcript.ai/mcp?key=YOUR_KEY as a remote Streamable HTTP server. The key can go in the URL, as here, or in an Authorization: Bearer header.

Try these prompts

Writing your own script?

The same key works over plain HTTPS from Python, JavaScript, n8n or Make. See the YouTube Transcript API.

Prefer a web tool?

No AI client needed — paste any link into the free web extractor.