YouTube MCP server
Bring YouTube
into your chat.
YouTube Video Transcript lets ChatGPT, Claude, and Grok find videos, fetch transcripts and comments, and work across a whole collection. Ask your question where you already work.
Connect your chat30 free credits every 30 days. No card required. Your chat’s connector requirements apply.
- Transcripts delivered through MCP
- 496
- Accounts using MCP
- 47
As of September 2026.
Start with a question.
You don’t need a link.
Search across YouTube or within one channel. Browse a creator’s playlists, check caption languages, and choose a source before spending a credit.
TRY ASKING
“Find MKBHD’s full iPhone reviews and check which have English captions. Show me the shortlist before fetching any transcripts.”
From finding a video to reading it: a recorded run.
28 ranked results
Searching @mkbhd for “iphone review” returned video titles and durations. The full review was result 21, not the first match.
search_channel_videos0 credits4 caption tracks
The selected review listed English, Hindi, and Spanish tracks. Its details also supplied the duration and publish date.
get_video_info0 creditsTranscript + 20 comments
With include_comments enabled, one call returned the transcript and the first Top comments page for the model to compare.
get_transcript2 credits at current pricing
Recorded workflow from 5 September 2026. Search returns YouTube’s ranking, not a curated shortlist. The three content calls above cost 2 credits at current pricing when the transcript and comments both succeed. See the recorded results.
What was said.
What people said back.
Bring the transcript and a first page of Top comments into the same call. Ask your model to compare the speaker’s claim with the reactions underneath it.
- A passage you can check.
- Request plain text for a summary, timestamps for finding a moment, or JSON for structured segments. Ask the model to cite its source and check important quotes against the video.
- Context before interpretation.
- The separate, free video-info tool supplies the title, channel, duration, view count, and caption tracks. Request dates for the publish timestamp, likes, and category when YouTube supplies them.
- Replies when the first take is not enough.
- Fetch more comments, switch to Newest, or open a reply thread with
get_video_comments. No transcript purchase is required. Each nonempty page costs one credit.
Top comments are a ranked sample, not a survey of everyone who watched. Ask the model to separate the creator’s claims, viewers’ opinions, and its own interpretation.
RECORDED RESULT EXCERPTS
Marques Brownlee
- Duration
- 22:21
- Published
- 2022-09-14
FROM THE TRANSCRIPT
“will be one of the most copied features”
FROM A TOP COMMENT · 32K LIKES
“just another one of the small marketing schemes that Apple does that works so well”
Bulk transcripts · paid plans
Compare transcripts across a collection.
Start a channel or playlist transcript job from the chat. Return later to compare recurring themes, find conflicting claims, or trace a topic across the videos you collected.
“Read the completed transcripts in my latest job. Compare what the videos say about battery life, cite each source, and tell me which videos you could not read.”
Already started a transcript job on the website? Ask the model to find it with list_jobs. Saved transcripts and comment samples can be read in batches without spending again.
Download the finished ZIP from the returned web link, signed in with the same Google account. Cancel from the website if needed. Job results are retained for 100 days.
See a collection read across 29 transcripts- 1
Check the balance.
See available credits before choosing a channel or playlist.
get_usage - 2
Start the collection.
Receive a job ID, its size, and a web link. Transcripts are fetched in the background.
create_bulk_job - 3
Check progress when needed.
See completed and failed videos. You do not need to keep the chat open while the job runs.
get_job - 4
Read what you already paid for.
Bring saved transcripts and optional comment samples back into chat without another fetch charge.
list_job_videos
MCP job caps: Starter 500 videos, Pro 2,000, Business 10,000. Channel jobs exclude Shorts and are sized to available credits within those caps; playlists must fit the plan cap and balance. Ask for coverage and failures before drawing a channel-wide conclusion.
Look around for free.
Pay for what you fetch.
Use the same balance in the website, API, and MCP. There is no extra YouTube Video Transcript subscription just to connect another chat client.
Comments are optional, not added automatically. A page usually holds about 20 top-level comments; reply pages can be smaller. Failed transcript fetches and empty or failed comments pages cost nothing.
Bulk transcript jobs reserve one credit per video, or two with a Top comments sample. Failed items and unused reservations are refunded. A comments failure never discards a successful transcript.
Connect your chat30 free credits every 30 days with Google sign-in. Paid plans start at $9/month for 1,000 credits. Compare plans.
| Action | Cost |
|---|---|
| Search, listings, video details, balance, and job status | Free |
| One successful transcript fetch | 1 credit |
| One nonempty comments page, including a reply page | 1 credit |
| Transcript with its optional nonempty Top comments page | 2 credits |
| Read saved transcripts or comment samples from a bulk job | Free |
Ask the model to check get_usage before a large fetch. Your AI provider’s subscription and usage charges are separate.
Connect your chat.
Choose your client below. Our server supports account sign-in through OAuth and API-key authentication for clients that accept a Bearer header. The guide shows which to use.
https://api.youtubevideotranscript.io/mcpHosted Streamable HTTP. No local server to install. Need a key? Get one from your account.
Claude
Connect Claude on the web or Desktop through the sign-in and consent flow. The guide also covers Claude Code and IDE setup with an API key.
Claude Free supports one custom connector. Team and Enterprise setup may need an organization owner.
Connect Claude Official connector requirementsChatGPT
Add our remote server as a custom app in Developer Mode, authorize your account, and select it in the conversation.
Developer Mode is available on the web for Plus, Pro, Business, Enterprise, and Education accounts. Workspace controls may apply.
Connect ChatGPT Official connector requirementsGrok
Add a custom MCP connector at grok.com/connectors, enter our server URL, and complete authentication. Grok discovers the tools for your conversations.
Business and Enterprise users may need their team admin to make the connector available.
Connect Grok Official connector requirements
Client access and organization policies can change. For Claude Code, IDE configuration, or direct model API calls, use the integration guides.
FIRST PROMPT AFTER CONNECTING · NO CREDITS SPENT
“Using YouTube Video Transcript, show my remaining credits and search YouTube for beginner sourdough tutorials. List candidates only. Don’t fetch transcripts or comments yet.”
Explore all 14 tools
- Find and scope
search_youtubesearch_channel_videoslist_channel_videoslist_playlist_videoslist_channel_playlistsget_channel_video_count- Check a video
get_video_info- Read the transcript
get_transcript- Read the conversation
get_video_comments- Work with a collection
create_bulk_joblist_jobsget_joblist_job_videos- Check your balance
get_usage
Before your first question.
What does the MCP connector actually do?
MCP, the Model Context Protocol, lets your AI client call external tools. YouTube Video Transcript supplies search results, video details, captions, comments, and bulk job results. Your chosen AI client uses that material to answer your question. This is a hosted connector: you do not need to run a YouTube scraper or MCP server yourself.
Do I need a video link before I start?
No. Ask the connector to search YouTube by topic or search within a specific channel. It can also list channel videos and playlists. Search and listings do not spend credits. Ask for a shortlist first, then choose which transcripts or comments to retrieve.
Can I use it on the free plan?
Yes. YouTube Video Transcript gives you 30 free credits every 30 days after Google sign-in, with no card required. Use them for individual transcripts and comments; search and video details are free. Creating a bulk transcript job through MCP requires a paid YouTube Video Transcript plan. Your chat provider has its own plan and connector requirements, separate from ours.
Does it work when a video has no captions?
Video details and accessible comments can still be useful without captions. The free get_video_info tool lists the caption tracks returned for the video before you request a transcript. Transcript retrieval needs an available track; it does not transcribe the audio from scratch. Failed transcript fetches are not charged, and the connector does not bypass private or restricted access.
Will asking a follow-up spend another credit?
Not if the model answers from material already in the conversation. A new transcript fetch costs another credit, even for the same video; a new nonempty comments page also costs one. Reading saved results through list_job_videos is free. Your AI provider’s own message, token, and context limits still apply.
Can it analyze an entire channel in one answer?
It can create a bulk transcript job and read its results in pages, but collection size and AI context limits still matter. MCP channel jobs cover long-form videos up to the plan cap and available credits. Some videos may have no captions. Ask the model to report the job’s coverage and failures, read results in batches, and cite the videos behind its conclusions. A collected channel does not mean every transcript fits into one answer.
Can I collect all comments through MCP?
The get_video_comments tool can page through available top-level comments and individual reply threads. Each nonempty page costs one credit, and this depends on the chat making further tool calls. For unattended comments collection with a hard credit limit and a JSON/CSV ZIP, use the website’s standalone Comments tool. The MCP bulk-job tool creates transcript collections, with an optional single Top comments page per successful transcript.
Collecting comments as a standalone dataset? Explore the Comments collector.