“What do reviewers praise and criticize in our category?”
Themes drawn from video transcripts, each with the videos behind it and the lines that prove it.
Research YouTube from Claude or ChatGPT with the buzzabout MCP.
Ask Claude or ChatGPT what YouTube reviewers and their viewers say about your product or your category. buzzabout finds the relevant videos, reads them through their captions, reads the comments and hands back the answer with sources.
Free trial with 1,000 credits. Copy page as Markdown Open in Claude
What do YouTube reviewers complain about most with robot vacuums?
buzzaboutaskqueryrender
I read 150 recent robot vacuum reviews through their captions, plus their comments. The top complaint is long hair wrapping the brush roll ("I cut hair off this thing every Sunday"), then missed corners and app connection drops, and the question viewers ask most is whether it copes with two shedding dogs. Want the reviews behind each complaint, ranked by views?
3.5 minmedian run time to fully analyzed posts
99%of research runs complete
7platforms in one connection
$49to start, MCP included
Questions you can ask about YouTube
Ask in plain English. Get answers with sources.
“What do viewers ask under the top videos on our topic?”
Questions from up to 100 comments per video, grouped into themes and ranked by likes.
“Which brands come up when people review our category?”
Brands and products named or shown in each video, counted across every review.
“How do the most-viewed videos on our topic open?”
Hook type and the exact opening line, next to real views and engagement rate.
“Which topics are gaining ground on YouTube in our space?”
Topics counted by week across titles, descriptions and transcripts, so the rising ones stand out.
“How do viewers feel about our product since launch?”
Sentiment on every video and its comments, charted over time.
What your AI gets back
Every YouTube post arrives transcribed, labeled and ready to use. Fifteen MCP tools do the work.
{
"source": "youtube",
"url": "https://www.youtube.com/watch?v=a1B2...",
"title": "Six months, two dogs, one robot vacuum",
"text": "Long-term review with pet hair tests.",
"summary": "Six-month review of a robot vacuum in a dog home.",
"hooks": [
{
"type": "Curiosity / Information Gap",
"text": "Six months ago I let a robot clean up"
}
],
"content_topics": [
{
"text": "pet hair cleaning"
},
{
"text": "robot vacuum upkeep"
}
],
"questions": [
{
"text": "Does the brush tangle with long dog hair?"
}
],
"sentiment": {
"positive": "0.55",
"neutral": "0.30",
"negative": "0.15"
},
"emotions": {
"joy": "0.45",
"surprise": "0.20",
"neutral": "0.35"
},
"tone_of_voice": {
"text": "authoritative",
"evidence": "I measured the bin after every run"
},
"narrative_structure": {
"text": "before/after",
"evidence": "Here is the floor on day one"
},
"cta": {
"text": "comment/engage",
"evidence": "Tell me your floor type below"
},
"media": [
{
"type": "video",
"transcription": "Six months ago I let a robot clean up after...",
"description": "Floor tests with dog hair, brush roll close-ups"
}
],
"created_at": 1793059200,
"dataset_run_at": 1793664000,
"content_analyzed_at": 1793664300,
"author": {
"title": "Example Channel"
},
"num_views": 48200,
"are_views_estimated": false,
"num_likes": 1930,
"num_comments": 214,
"num_shares": 0,
"engagement_rate": "0.044",
"category": "technology_and_computing",
"vibe": {
"raw_authentic": 0.6,
"clinical": 0.4
},
"language": "en"
}15 MCP tools
Research
askAsk a research question in plain English; get a cited answer.get_messageRead the research agent's answer as it lands.
Collect
collect_mentionsPreview a topic, then collect and analyze the posts.ingest_mentionSave and analyze one public post by URL.get_runFollow any run's progress, with long-polling.get_save_mention_statusTrack a saved post's analysis.
Analyze
queryAsk your data anything: counts, themes, trends, comments, audience.list_mentionsRead collected posts with all their AI fields.detect_patternsGroup posts into 2 to 15 named themes.collect_audienceProfile the people commenting, in aggregate (Pro and up). Not available for YouTube.renderShow answers as interactive charts and tables.
Track and manage
list_tracking_agentsSee your scheduled topic trackers.get_tracking_agentRead one tracker and what it found.list_collectionsSee everything you have collected.get_meCheck your plan and credit balance.
Every field in a video
Transcript and visuals
media[].transcriptionThe spoken words, from the video's caption track.media[].descriptionWhat the video shows, described by AI from its preview frames.languageThe video's language, as an ISO 639-1 code.
Hooks and format
hooksHook type plus the exact opening line.narrative_structureStory shape, with the quote that shows it.ctaThe call to action, with evidence.content_intentionInform, entertain, sell or engage.categoryContent category, such astechnology_and_computing.vibeTwelve mood scores, from raw and authentic to corporate.
Topics and questions
summaryA short summary of the video.content_topicsThe topics it covers.questionsQuestions asked in the video.mentioned_brandsBrands and products named or shown.entitiesPeople, places, events and organizations named; brands are inmentioned_brands.
Sentiment and tone
sentimentPositive, neutral and negative shares.sentiment_scoreOne number from negative to positive.emotionsJoy, surprise, anger, sadness, fear, disgust, neutral.tone_of_voiceTone label with the line that sets it.
Engagement
num_viewsReal view count from YouTube.num_likesLikes.num_commentsComment count.engagement_rate(likes + comments) / views, computed for you.
Comments
comments[].bodyThe comment text.comments[].num_likesLikes on the comment.comments[].num_repliesHow many replies it drew.comments[].created_atWhen it was posted.comments[].sentimentHow each comment feels, scored (viaquery).
Read comments through MCP query or buzzabout__ask. They are not in list_mentions or the REST response yet.
Also in every record: url, title, text, author.title, author.avatar, media[].cover_image, created_at, dataset_run_at, content_analyzed_at.
How it compares
Other data providers give you raw posts. buzzabout gives you answers.
| Capability | buzzabout | Other data providers |
|---|---|---|
| What you get back | YesPosts already read, labeled and grouped into themes | NoRaw posts |
| Works in Claude, ChatGPT and other AI agents | YesOne MCP connection, every plan | YesRaw-data MCP servers |
| All 7 platforms in one request | YesOne question covers all 7 | PartlyOne scraper per platform |
| Off-topic posts filtered, never billed | YesAI check on every post | NoYou pay for every record returned |
| Hooks, topics, sentiment and tone | YesOn every post | NoBring your own AI |
| Comments with every post | YesUp to 100 per post, included | PartlySeparate scraper, billed per comment |
| Themes across many posts | Yes2 to 15 named themes per question | NoNone |
| Price | Starting at $49, everything included | From about $472 a month for the full stack |
Everything included, starting at $49
Bought separately, the same stack costs about $472 a month, before anyone builds anything.
Bought separately$472a month
buzzabout$49Starter, first month
What buzzabout includes
- DataPosts, comments and transcripts from 7 platforms
- ProcessingHooks, pains, formats and sentiment on every post
- Research toolsThemes, audiences, trends and briefs
- StorageEvery collection kept and shareable by link
- WorkflowsListening agents that run on a schedule
- MemoryYour past research, ready for the next question
- IntegrationsClaude, ChatGPT, Slack alerts and the API
Two ways to use it
MCP on every plan. REST API on Business.
In Claude or ChatGPT
Add buzzabout as a connector and ask in plain English. That's it.
- Start your free trial.
- In Claude or ChatGPT, add a custom connector with
https://mcp.buzzabout.ai/mcpand sign in. - Ask about YouTube.
In Claude Code or your own agent
Install the MCP server and let Claude Code build on it: digests, dashboards, alerts.
- Run
claude mcp add --transport http buzzabout https://mcp.buzzabout.ai/mcp. - Tell Claude Code what to build, for example a weekly digest of YouTube posts about your competitors.
- Schedule it with tracking agents and webhooks. The REST API is there on Business.
1. In Claude, add a custom connector.
2. URL: https://mcp.buzzabout.ai/mcp
3. Sign in, then ask:
"What do people on YouTube say about
robot vacuum reviews for pet hair?"1. In ChatGPT, add a custom connector (MCP server).
2. URL: https://mcp.buzzabout.ai/mcp
3. Sign in, then ask:
"What do people on YouTube say about
robot vacuum reviews for pet hair?"claude mcp add --transport http buzzabout https://mcp.buzzabout.ai/mcp
# then, in Claude Code:
> Build a weekly digest of YouTube posts about
robot vacuum reviews for pet hair, with the top hooks, questions
and sentiment, posted to our Slack webhook.{
"mcpServers": {
"buzzabout": {
"url": "https://mcp.buzzabout.ai/mcp"
}
}
}curl -X POST https://api.buzzabout.ai/v1/ask \
-H "x-api-key: $BUZZABOUT_API_KEY" \
-H "Content-Type: application/json" \
-d '{"message": "What do people on YouTube say about robot vacuum reviews for pet hair?"}'Pricing
Every plan includes the MCP and the full analysis.
Starter
$49first month, then $98/mo
2,000 credits a month. Up to 250 posts per research.
Try for freePro
$178/month
5,000 credits a month. Audience research. Up to 1,000 posts per research.
Try for freeBusiness
$450/month
20,000 credits a month. REST API. Unlimited listening agents.
Try for freeFree trial with 1,000 credits. A card is needed to sign up. See pricing.
Questions
Is the YouTube Data API free?
Free, but rationed. The default quota is 10,000 units a day and each search costs 100 units, so about 100 searches a day. You cannot buy more quota; an extension needs an audit. buzzabout has no quota units: you pay one credit per video you keep, analysis included, and you ask from Claude in plain English. Try it free.
Can I get YouTube transcripts through an API?
Not from the official Data API. buzzabout pulls the transcript from each video's caption track when it has one, up to 18,000 characters. buzzabout__ask switches this on; with collect_mentions, set enable_transcribing: true, up to 18,000 characters, and hands it to Claude with a summary, topics, questions and sentiment, at no extra cost.
Can I get YouTube comments through an API?
Yes. The official API returns them through commentThreads, inside your daily quota. buzzabout collects up to 100 top-level comments per video at no extra cost, with likes, reply counts, timestamps and sentiment. Ask Claude or ChatGPT about them in plain English, or query them through the MCP.
What is the best YouTube MCP for research?
buzzabout. Connect it to Claude or ChatGPT and YouTube videos come back read through their caption transcripts, described and labeled for topics, questions, hooks and sentiment, with comments, through one connection that covers six other platforms. A typical run is ready in 3.5 minutes, 99% of runs complete, and you pay only for videos you keep.
Why not just use a scraper or another data provider for YouTube?
A scraper hands you raw YouTube records and, at best, the caption file, and stops. Relevance filtering, AI labels, themes and storage are separate tools you buy, connect and maintain. Add themes, scheduled runs, memory and integrations and the full stack starts at about $472 a month, before anyone builds anything. buzzabout includes all of it, starting at $49, through one connection your agent already speaks.
How much does buzzabout cost for YouTube research?
Starting at $49 for your first month on Starter, then $98 a month. Each post you keep costs one credit, with the analysis included. See pricing.
Does it work in ChatGPT?
Yes. Add buzzabout as a custom connector in ChatGPT with https://mcp.buzzabout.ai/mcp and sign in. Custom connectors depend on your ChatGPT plan.
Do I pay for off-topic or duplicate YouTube posts?
No. You pay only for the posts you keep.
Do I need to code?
No. Add buzzabout to Claude or ChatGPT and ask in plain English. Growth engineers can use the same MCP in Claude Code or Cursor.
How do I start?
Start a free trial with 1,000 credits, add buzzabout to Claude or ChatGPT and ask your first question.
Try it on your own topic
Free trial. Works in Claude and ChatGPT.
Free trial with 1,000 credits · MCP on every plan