Skip to content
Get transcript

YouTube transcript MCP server

Read a YouTube transcript inside Claude, Cursor or Codex. Keep the source captions and timestamps.

https://scribiz.com/mcp
Create your API keySet up your client

Scribiz sign-in and API key required. Each account starts with 30 caption credits, once.

Connect your AI assistant

Create a key with MCP or All access in your dashboard. Choose your client. Set SCRIBIZ_API_KEY in the environment that starts it and keep the key out of project files.

Set your key in the environment that starts your client
export SCRIBIZ_API_KEY="YOUR_SCRIBIZ_API_KEY"

Run this in your terminal, then open /mcp in Claude Code to check the connection.

Terminal
claude mcp add --transport http scribiz https://scribiz.com/mcp \  --header "Authorization: Bearer $SCRIBIZ_API_KEY"

Official Claude Code MCP setup

Get a YouTube transcript from a link

Paste a YouTube URL into your assistant and ask it to call get_transcript.

Prompt for your AI assistant
Use Scribiz get_transcript for this YouTube video: YOUR_YOUTUBE_URL. Return txt with the source timestamps and keep the original caption language.

Replace YOUR_YOUTUBE_URL with your video link. Your assistant receives the existing captions. It can then summarize the returned text, make notes or quote a passage with its source timestamp.

Source captions, timestamps and formats

Choose a format and an available caption language. The timestamps come from the source caption track.

YouTube get_transcript options
OptionWhat to request
formattxt (default), md, srt, vtt or json. TXT includes timestamp markers.
languageAn available source track, such as en or pt-BR. This selects captions; it does not translate.
from / toRead a time range, for example 1:00 to 2:30.
cursorContinue with nextCursor from the result and keep the other arguments the same.
max_charsSet the page size from 1,000 to 150,000 characters.
Actual stored response metadata: 7 Oct 2026
{  "captionCreditsCharged": 0,  "durationSeconds": 3610,  "format": "txt",  "language": "uk",  "minutesUsed": 0,  "origin": "captions-auto",  "status": "done",  "timing": "caption",  "videoId": "youtube:1U-x9bdnwx0"}

A recorded read of this YouTube video. The caption language is Ukrainian. This transcript was already stored, so the read used zero caption credits. A new successful transcript uses one.

Frequently asked questions

Sign in to Scribiz and create an API key with MCP or All access. Set it as SCRIBIZ_API_KEY for your client and follow the setup steps. You do not need a Google or YouTube API key.

Each Scribiz account receives 30 caption credits once. One new successful transcript uses one credit, regardless of video length. Stored reads, failed requests and re-exports use zero credits. Caption credits are separate from transcription minutes.

New caption retrieval has a daily limit. If you reach it, retry after midnight UTC. Stored transcripts remain available. Failed requests use no caption credits.

The request returns a missing-captions error and uses no caption credits. This transcript tool retrieves existing YouTube captions. It does not switch to audio transcription or a model reading the video.

Scribiz keeps the source caption text and cue timestamps, including automatic captions. It does not proofread or rewrite them. The result identifies the caption source when known. Caption errors from YouTube remain in the transcript.

Set language to a BCP-47 code such as en or pt-BR to request an available caption track. An unavailable language returns an error. This option selects source captions; it does not translate them.

TXT includes timestamp markers. SRT and VTT use subtitle cue times, and JSON keeps structured timing. Use from and to to read a time range. For a long transcript, continue with nextCursor while keeping the other arguments the same.

Read the caption limits and error codes.

More ways to read video transcripts

Bulk YouTube transcripts

Have more than one video? Use the separate bulk transcript tool.

Video MCP for other sources

For uploaded videos, TikTok and Vimeo, see the general video MCP guide.

Read the transcript in your assistant

Create a Scribiz key, connect your client and paste one YouTube link.

Create your API key