How to Get YouTube Transcripts in n8n

Set up the credential once

Create a free API key at transcript-yt.com. In n8n, add a Header Auth credential with the name Authorization and the value Bearer ts_live_... so the key is stored once and never pasted into individual nodes.

Get one transcript with an HTTP Request node

Add an HTTP Request node with method GET and URL https://transcript-yt.com/v1/transcript. Set Authentication to your Header Auth credential, then add a query parameter url with the video link, for example from a previous node: {{ $json.videoUrl }}.

The response is JSON with title, channel, language, segments, and the full text, so the next node can use {{ $json.text }} directly. Add a query parameter format=text if you only want plain text, which is the easiest input for an OpenAI or Anthropic node.

• Method: GET
• URL: https://transcript-yt.com/v1/transcript
• Query parameters: url = {{ $json.videoUrl }}, optional language, translate_to, or format
• Timeout: set to 120 seconds under Options, since long videos without captions are transcribed from audio

Process a whole channel or playlist

Use a first HTTP Request node on GET /v1/playlist with url set to a channel @handle or playlist link. It is free and returns a videos array. Add a Split Out node on videos, then the transcript HTTP Request node with url = {{ $json.videoId }}. n8n runs the request once per video.

To save executions, send up to 50 IDs at once to POST /v1/transcripts instead, with a JSON body of {"urls": [...]}, and split the results array afterwards.

Let an n8n AI Agent read YouTube videos

n8n's MCP Client Tool node connects an AI Agent to a remote MCP server. Point it at https://transcript-yt.com/mcp using Header Auth with your key. The agent then gets get_transcript, search_youtube, list_playlist_videos, and list_languages as tools, and decides on its own when to fetch a transcript.

That makes chat-style workflows easy: a Telegram or Slack trigger, an AI Agent with the MCP Client Tool attached, and users can send a YouTube link and ask for a summary.

Handle failures in the workflow

Turn on Continue On Fail for the transcript node, or use the error output, so one bad video does not stop a batch. Errors come back as {"error": {"code", "message"}}. Retry BLOCKED, RATE_LIMITED, and UPSTREAM_ERROR; treat VIDEO_UNAVAILABLE and NO_CAPTIONS as final. Failed requests are never billed.