YouTube Transcript MCP By GetYouTubeText

official

Fetch YouTube transcripts at an ultra fast speed

What can you do with YouTube Transcript By GetYouTubeText MCP?

  • Fetch a video transcript — Ask for the transcript of any YouTube link or video ID, with optional language selection and time-range filtering.
  • Get timestamped or paragraph transcripts — Request timestamps on each line or merged paragraph chunks, and search within the transcript for specific words.
  • Look up video details — Retrieve a video's title, channel, and thumbnail without pulling the full transcript.
  • List a channel's latest uploads — Get the 15 newest videos from a channel by ID, handle, or channel URL.

Hosted MCP Server

npx add-mcp 'https://getyoutubetext.com/mcp'

Installs into Claude Code, Codex, Cursor and more

Documentation

YouTube Transcript API documentation

Everything is a GET request that returns JSON, unless you ask for a file format. The base URL is https://getyoutubetext.com.

Quickstart

Create an account, then send your first request. New accounts start with 100 free credits.

curl "https://getyoutubetext.com/v1/transcript?url=jNQXAC9IVRw" \
  -H "Authorization: Bearer YOUR_API_KEY"
{
  "video_id": "jNQXAC9IVRw",
  "title": "Me at the zoo",
  "author": "jawed",
  "author_url": "https://www.youtube.com/@jawed",
  "language": "en",
  "available_languages": ["en", "de"],
  "duration_seconds": 19,
  "cached": true,
  "segment_count": 6,
  "text": "All right, so here we are, in front of the elephants...",
  "segments": [
    { "text": "All right, so here we are, in front of the elephants", "start": 1.2, "duration": 2.16 },
    { "text": "the cool thing about these guys is that they have really...", "start": 5.318, "duration": 2.656 }
  ]
}

Authentication

Send your key in the Authorization header on every request (an X-API-Key header also works). Keys start with gyt_live_, never expire, and can be revoked from the dashboard. An account can have up to 10 keys.

Authorization: Bearer gyt_live_…

Keep keys on your server. A key in browser code or a public repo can be used by anyone who finds it.

Credits

  • Every successful request costs 1 credit: a transcript (whatever the video length or format), video info, or a channel’s latest uploads.
  • Failed requests cost nothing: no captions, private or missing videos, bad input, rate limits and our own errors. Checking your account (/v1/me) is free too.
  • Paid plan credits refill every month, on yearly plans too. Unused plan credits do not carry over. The free plan’s 100 credits are given once.
  • On a paid plan you can buy credit packs: 5,000 credits that never expire, used after your plan credits run out.
  • Repeat requests for the same video are served from cache. They are faster and still cost 1 credit.

Once your key is accepted, every transcript, video and channel response, failed ones included, carries X-Credits-Charged and X-Credits-Remaining headers, so you can track spend without another call.

Transcript 1 credit

GET /v1/transcript

ParamTypeDescription
urlstring, requiredA YouTube URL (watch, youtu.be, shorts, embed, live) or a bare 11-character video ID.
langstringLanguage code such as es or pt-BR. Defaults to English, then the video’s original track. If the video has no track in that language you get the default one, and language in the response tells you which.
formatstringjson (default), text, srt, vtt, csv or md.
timestampsbooleanSet true to prefix each line with [HH:MM:SS] in text output and in the text field of JSON. In md output each line starts with a clickable [MM:SS] link to that moment of the video.
paragraphsbooleanSet true to merge caption lines into paragraph-sized chunks, split on pauses and sentence ends.
chunk_sizeintegerWith paragraphs=true, the most characters per paragraph. 50 to 10000, default 400.
start / endnumberSeconds. Return only the part of the transcript inside this range.
searchstringReturn only the lines containing this text, case-insensitive. Combine with timestamps=true to see where each was said.
segmentsbooleanSet false to drop the segments array and return only text (JSON only).

Response fields (JSON)

  • language: the track you got. available_languages: every track the video has.
  • duration_seconds, segment_count, cached
  • text: the whole transcript as one string
  • segments: array of { text, start, duration } with times in seconds

Examples

curl "https://getyoutubetext.com/v1/transcript?url=dQw4w9WgXcQ&format=srt" \
  -H "Authorization: Bearer YOUR_API_KEY" -o captions.srt

Video info 1 credit

Title, channel and thumbnail for a video, without the transcript. Useful for showing what a link is. A missing or private video returns VIDEO_UNAVAILABLE and costs nothing.

{
  "video_id": "jNQXAC9IVRw",
  "url": "https://www.youtube.com/watch?v=jNQXAC9IVRw",
  "title": "Me at the zoo",
  "author": "jawed",
  "author_url": "https://www.youtube.com/@jawed",
  "thumbnail_url": "https://i.ytimg.com/vi/jNQXAC9IVRw/hqdefault.jpg"
}

Channel latest 1 credit

The 15 newest uploads of a channel, newest first. channel takes a channel ID (UC…), an @handle or a channel URL (youtube.com/@handle or youtube.com/channel/UC…). Older /c/ and /user/ links are not supported, so use the handle instead. Poll it to catch new videos, then fetch their transcripts.

{
  "channel_id": "UCBJycsmduvYEL83R_U4JriQ",
  "channel": "Marques Brownlee",
  "channel_url": "https://www.youtube.com/channel/UCBJycsmduvYEL83R_U4JriQ",
  "video_count": 15,
  "videos": [
    {
      "video_id": "hdMszZ2ah5Q",
      "url": "https://www.youtube.com/watch?v=hdMszZ2ah5Q",
      "title": "…",
      "published": "2026-11-11T18:43:56+00:00",
      "views": 5319666,
      "thumbnail_url": "https://i.ytimg.com/vi/hdMszZ2ah5Q/hqdefault.jpg"
    }
  ]
}

Account Free

Plan, credits left and rate limit for the key you send. billing is month, year or null on the free plan. Free, and it does not count toward your rate limit.

{
  "plan": "free",
  "billing": null,
  "credits_remaining": 96,
  "plan_credits_remaining": 96,
  "plan_credits_total": 100,
  "topup_credits": 0,
  "period_end": null,
  "rate_limit_per_minute": 30
}

Errors

Errors always have the same shape. Branch on code, which is stable. The message is written for people and may change.

{
  "error": {
    "code": "NO_CAPTIONS",
    "message": "This video has no captions, so there is no transcript to return.",
    "retryable": false
  }
}
StatusCodeMeaningRetry?
400INVALID_URLThe url is missing or is not a YouTube link or video ID.No
400INVALID_PARAMA parameter has a value we cannot use. The message says which.No
401INVALID_API_KEYThe key is missing, wrong or revoked.No
402INSUFFICIENT_CREDITSYou are out of credits.No
404NO_CAPTIONSThe video exists but has no captions.No
404VIDEO_UNAVAILABLEThe video is private, removed, age-restricted or does not exist.No
404CHANNEL_NOT_FOUNDNo channel matches that ID or handle.No
429RATE_LIMITEDToo many requests this minute. Wait for the Retry-After header.Yes
502UPSTREAM_ERRORYouTube did not answer in time. Retry after a second or two.Yes
500INTERNAL_ERRORSomething broke on our side.Yes

Rate limits

Limits are per account, shared by all your keys, the MCP server and the dashboard playground, counted over a rolling minute. Transcript, video and channel responses include X-RateLimit-Limit and X-RateLimit-Remaining. Past the limit you get a 429 with a Retry-After header in seconds. Yearly plans get a higher limit; on Scale yearly, ask us for more.

PlanMonthly billingYearly billing
Free3030
Starter120200
Pro300500
Scale6001000

MCP server

A hosted Model Context Protocol server lives at https://getyoutubetext.com/mcp (Streamable HTTP). It uses the same credits as the REST API.

Sign in from the app (ChatGPT, claude.ai)

Apps with a Connect button only need the server URL. They open a GetYouTubeText sign-in page, you approve the connection, and the app receives its own token. This is OAuth 2.1 with PKCE and dynamic client registration; the metadata is at /.well-known/oauth-authorization-server. Connected apps can be removed from the MCP page in your dashboard.

ChatGPT: open Plugins, click Add, then Add custom MCP server, and paste the server URL with OAuth. Under Advanced OAuth settings, set Registration method to User-Defined OAuth Client and enter the client ID getyoutubetext-chatgpt, with no secret and the auth method none.

Use an API key (Claude Code, Cursor, VS Code)

claude mcp add --transport http getyoutubetext https://getyoutubetext.com/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"
ToolArgumentsCost
get_transcripturl (required), lang, timestamps, paragraphs, start, end, search1 credit
get_video_infourl (required)1 credit
get_channel_latest_videoschannel (required)1 credit

For get_transcript, timestamps and paragraphs default to true, because models read timestamped paragraphs better. The REST API defaults both to false.

Tool errors come back as a normal tool result with isError: true and a message starting with the same error code as the REST API, so the model can read it and decide what to do.