How to Connect Claude with YouTube Data
TL;DR
- Claude doesn't have access to YouTube data. Ask about any channel and Claude will give you a confident answer built on numbers it invented.
- TubeLab's YouTube MCP server connects Claude to live YouTube data. 400K+ searchable channels, millions of verified outlier videos, full video transcripts, and top comments.
- Setup takes 4 steps, no coding. Go to claude.ai/settings/connectors, add TubeLab, log in, approve. Done.
- Several tools are free. Channel lookups, video details, transcripts, and comments cost nothing.
- Claude is the only AI that supports TubeLab's pre-built prompt workflows natively. Click the + menu to run Channel Roasts, Niche Analysis, and Trend Reports without typing anything.
- Connect TubeLab to Claude
I asked Claude to help me evaluate the finance niche for a new YouTube channel. Without TubeLab connected, Claude said the niche was "growing steadily with moderate competition" and estimated that channels in the 10K-50K subscriber range "typically see 15,000-30,000 views per video." It named three channels as examples. Sounded reasonable.
Then I connected TubeLab's YouTube MCP server and asked the same question.
Claude searched TubeLab's database of 400K+ channels, filtered for finance channels with 10K-50K subscribers, and found 380+ monetized channels in that range. "Moderate competition" was wrong. This was one of the most crowded niches on the platform. Median views were 4,800 per video, not the 15,000-30,000 Claude had estimated. Two of the three channels it previously named didn't exist. The third had 7,200 subscribers, not the "roughly 35K" Claude had claimed.
That's the difference. One answer sounds right. The other is right. And you can't tell which is which until you connect Claude to the actual data.
This isn't a flaw in Claude. It's the strongest reasoning tool available for analyzing data, spotting patterns, and generating strategy. The problem is that without a live data connection, it has no YouTube data to reason about. It fills the gap with plausible-sounding estimates that feel authoritative because Claude's writing is authoritative. The analysis is real. The data underneath it isn't.
TubeLab's YouTube MCP gives Claude the data. If you already use TubeLab's niche finder or outlier finder on the web, the MCP puts that same dataset inside your Claude conversations. No more switching tabs. No more copy-pasting data. Ask Claude and it pulls from TubeLab automatically.
Connect TubeLab to Claude (4 Steps)
No API keys, no terminal, no code. The whole thing is a browser login.
Step 1. Go to claude.ai/settings/connectors and click Add custom connector.
Step 2. Set the name to TubeLab and paste the MCP server URL:
https://public-api.tubelab.net/mcp
Step 3. Click Connect. Sign in to your TubeLab account (or create one) and click Approve.
Step 4. Click Configure on the TubeLab connector and toggle "Always allow" so Claude can call TubeLab's tools without asking permission each time.
Open a new chat and ask anything about YouTube. Claude will call TubeLab's tools automatically.
What Claude Gets Access To
Once connected, Claude has YouTube data tools. It picks the right one based on your question. You just ask in plain language.
Free tools (no credits, no subscription): look up any channel's full profile (subscribers, views, upload cadence), pull any channel's recent videos or Shorts with view counts and engagement data, get full metadata on any video, pull complete transcripts from any video's captions, and read the top comments on any video sorted by engagement.
Paid tools (credits from your TubeLab account): search 400K+ channels with 20+ filters including niche, subscriber range, language, and monetization status (5 credits), find channels similar to any channel you specify (5 credits), search millions of verified outlier videos by topic, views, and performance multiplier (2 credits), and find outliers from channels similar to a given channel (2 credits).
Claude-Only: Pre-Built Research Workflows
This is a feature only Claude has. TubeLab's YouTube MCP server includes prompt workflows built into the protocol, and Claude surfaces them natively in the + menu at the bottom of any chat. No copy-pasting. No typing prompts. Click a workflow and it runs.
- Channel Roast: Drop in any channel URL. Claude pulls data, analyzes performance, and delivers a strategic breakdown of what's working and what isn't.
- Niche Analysis: Specify a category. Claude maps the competitive space, subscriber ranges, and monetization rates.
- Niche Research: Goes deeper. Claude finds outlier videos, cross-references patterns, and surfaces specific content opportunities.
- Trend Report: Identifies topics gaining traction based on recent outlier activity in any niche.
- Video Ideation: Combines outlier data, comments, and transcripts to generate video ideas backed by proven audience demand.
How to Use Claude for YouTube Research (Not Just Lookups)
Pulling a channel's subscriber count is the simple use case. The reason connecting TubeLab to Claude matters is that Claude can chain multiple data pulls into a research workflow and analyze the results. That's where it stops being a lookup tool and starts being a research partner.
Here's how to get the most out of it.
Niche Validation: What to Actually Look For
Most creators validate a niche by browsing YouTube for an hour and deciding based on gut feeling. The problem isn't effort. It's that they don't know which data points actually predict whether a niche will work.
Three things matter when evaluating a niche: how many monetized channels exist in your target subscriber range (competitive density), what view counts look like at that subscriber level (audience demand), and whether channels are growing or stagnant (trajectory).
Here's how to get all three in one Claude conversation:
"Search for monetized YouTube channels in the home renovation niche with 10K-100K subscribers that upload at least weekly in English. How many exist? What's the median view count per video?"
Claude calls TubeLab's channel search and returns actual channels. Count matters: 200+ monetized channels means the niche supports the business model but competition is real. Under 20 means the audience might not be there, or the niche is too small to support monetization.
But count alone isn't enough. Ask Claude to go deeper:
"Pull the last 10 videos from the top 3 channels in those results. Are view counts consistent or do they spike and crash? Are the channels growing or plateaued?"
Consistent views mean a reliable audience. Spike-and-crash patterns mean the channel depends on viral hits and can't sustain watch time between them. This is the difference between a niche with a real audience and a niche that looks active from subscriber counts but doesn't generate steady watch time.
Before doing this in Claude, check TubeLab's niche finder on the web for a quick overview of saturation, CPM estimates, and growth trends. The niche finder tells you if the space is worth entering. Claude tells you exactly what's happening inside it.
Competitor Teardown: The Three-Layer Method
Surface-level competitor analysis looks at subscriber counts and view numbers. It tells you almost nothing useful. The insights come from going three layers deep: performance data, then content structure, then audience response.
Layer 1: Performance patterns. Start with the data.
"Get all the data on @competitor. Pull their last 30 videos. Which 5 performed best and which 5 performed worst? What separates them? Compare titles, durations, topics, and publish days."
Claude gets the channel overview and video list from TubeLab, then analyzes the full set. You might find that their best videos are all 12-15 minutes while their worst are over 20. Or that list-based titles outperform question-based titles. Or that Tuesday uploads consistently beat Friday uploads. These patterns exist in every channel's data. Most creators never see them because finding them manually takes hours.
Layer 2: Script structure. Go deeper into their best performer.
"Get the transcript from their #1 video. How long was the hook before the main content started? How did they structure transitions between sections? Where were the strongest moments in the script?"
Claude pulls the full transcript from TubeLab and analyzes the writing decisions that made the video work. You're not studying what they said. You're studying how they structured what they said: intro length, section pacing, where they placed their strongest points, and how they closed.
Layer 3: Audience response. See how viewers actually reacted.
"Now get the top comments on that video. What are viewers praising most? What questions did they ask that the video didn't answer? What follow-up topics did they request?"
This layer tells you what the audience wanted that they didn't get. The video's biggest gap is your biggest opportunity. The questions viewers asked are content ideas they've already validated by asking.
Three layers, three TubeLab tools, one Claude conversation. Channel data, transcript analysis, comment mining. What used to take an afternoon takes three prompts.
The free TubeLab Chrome Extension adds another layer while you browse YouTube: monetization status, RPM estimates, and performance data visible on any channel with one click.
Content Ideation: Building Ideas from Proof
The strongest video ideas don't come from brainstorming. They come from evidence. Here's a method that uses Claude and TubeLab to build ideas from two data sources: outlier patterns and audience gaps.
Start with outliers. Find what the audience has already proven they want.
"Find outlier videos about personal finance from the last 90 days with at least 10x their channel's average views. Show me the top 10. What topics keep appearing?"
Claude searches TubeLab's outlier database and identifies patterns across the viral videos. If three of the top 10 are about the same subtopic, the audience is hungry for that topic right now.
Then check if the trend is real or a one-off:
"Find outlier videos from channels similar to the top result. Are more creators seeing outsized performance on this subtopic?"
If multiple channels see outlier performance on the same topic, it's a confirmed trend. If it's isolated to one channel, it might not translate to your audience.
For visual browsing of outlier data, TubeLab's outlier finder on the web offers the same dataset in a point-and-click interface.
Then mine comments for gaps. Take the top 3 outlier videos and ask Claude:
"Get the top comments on these 3 videos: [URLs]. What questions appear across multiple comment sections? What did viewers say was missing? What did they disagree with?"
Questions that appear across multiple comment sections are high-confidence content ideas. The audience asked for them more than once, on different videos, from different creators.
Turn it into video concepts:
"Based on the outlier topics and the comment gaps, suggest 5 video ideas. For each, give me a working title, the audience question it answers, and a one-paragraph outline."
Every idea traces back to specific data: an outlier pattern that proved the topic works, and a comment gap that shows the audience wants more. No brainstorming in a vacuum. No hoping a topic lands.
Set Up a Claude Project for Ongoing YouTube Research
One-off research conversations are useful. But the creators who get the most from TubeLab's YouTube MCP treat YouTube research as an ongoing process, not a one-time task.
Claude's Projects feature lets you create a dedicated workspace for YouTube research with a custom system prompt that carries across every conversation in the project. Set one up with instructions like:
"You are my YouTube research assistant. When I ask about a niche, always check competitive density, view count ranges at my target subscriber level, and whether top channels are growing or stagnant. When I ask about a competitor, use the three-layer method: performance data first, then transcript analysis, then comment mining. Always use TubeLab's tools for any YouTube data."
Every conversation you start inside that Project follows those instructions. You don't need to re-explain your research methodology each time. Ask "how's the finance niche looking this month?" and Claude automatically runs the full validation process using TubeLab's tools because the Project instructions tell it to.
This also makes it easy to track changes over time. Run the same niche query monthly. Compare the channel counts, view averages, and outlier patterns. You'll see whether the niche is getting more saturated, whether view counts are rising or falling, and whether the content opportunities from last month are still open or already taken.
Frequently Asked Questions
Does Claude have YouTube data without TubeLab? No. Claude has general knowledge about YouTube but cannot access current channel stats, video performance, or niche data. Any specific numbers about YouTube are from outdated training data or invented. TubeLab gives Claude a live connection to YouTube.
What does it cost? Six tools are free with a free TubeLab account: channel lookups, video details, transcripts, and comments. The four search tools (channels, related channels, outliers, related outliers) consume credits and require a TubeLab subscription at $178.80/year.
Can I use the free tools without a paid subscription? Yes. Create a free TubeLab account, connect it to Claude, and use the six free tools immediately. Look up any channel, pull any video's details, read any transcript, and mine any video's comments at no cost.
What are the pre-built prompt workflows? Research templates built into TubeLab's MCP server that Claude surfaces in the + menu: Channel Roast, Niche Analysis, Niche Research, Trend Report, and Video Ideation. Each runs a multi-step research workflow. You can also browse and copy them from the prompt library.
Can Claude modify my YouTube channel? No. TubeLab's MCP is read-only. Claude can query publicly available data but cannot upload, edit, or change anything on any channel.
Does this work with Claude Projects? Yes. Set up a Project for YouTube research with custom instructions (like "always check niche data before making content recommendations") and Claude will follow those instructions while using TubeLab's tools throughout the conversation.
Sources
TubeLab - YouTube MCP for Claude - Product page and setup overview.
TubeLab - Claude Connector Setup Guide - Step-by-step documentation.
TubeLab - YouTube MCP Server Documentation - Full tool list, authentication, and credits.
TubeLab - Prompt Library - Pre-built research workflows.
Model Context Protocol - Introduction - Official MCP documentation.
Give Claude Real YouTube Data
Claude is the strongest reasoning tool available. It's wasted when the data it reasons about doesn't exist.
Connect TubeLab's YouTube MCP and every question about YouTube gets answered with verified data from actual channels. Six tools are free. Setup takes four steps.
For the full research toolkit (outlier discovery, channel search across 400K+ channels, niche analysis), TubeLab costs $178.80/year. No tiers, no feature gating, no upsells.
TL;DR
- Claude doesn't have access to YouTube data. Ask about any channel and Claude will give you a confident answer built on numbers it invented.
- TubeLab's YouTube MCP server connects Claude to live YouTube data. 400K+ searchable channels, millions of verified outlier videos, full video transcripts, and top comments.
- Setup takes 4 steps, no coding. Go to claude.ai/settings/connectors, add TubeLab, log in, approve. Done.
- Several tools are free. Channel lookups, video details, transcripts, and comments cost nothing.
- Claude is the only AI that supports TubeLab's pre-built prompt workflows natively. Click the + menu to run Channel Roasts, Niche Analysis, and Trend Reports without typing anything.
- Connect TubeLab to Claude
I asked Claude to help me evaluate the finance niche for a new YouTube channel. Without TubeLab connected, Claude said the niche was "growing steadily with moderate competition" and estimated that channels in the 10K-50K subscriber range "typically see 15,000-30,000 views per video." It named three channels as examples. Sounded reasonable.
Then I connected TubeLab's YouTube MCP server and asked the same question.
Claude searched TubeLab's database of 400K+ channels, filtered for finance channels with 10K-50K subscribers, and found 380+ monetized channels in that range. "Moderate competition" was wrong. This was one of the most crowded niches on the platform. Median views were 4,800 per video, not the 15,000-30,000 Claude had estimated. Two of the three channels it previously named didn't exist. The third had 7,200 subscribers, not the "roughly 35K" Claude had claimed.
That's the difference. One answer sounds right. The other is right. And you can't tell which is which until you connect Claude to the actual data.
This isn't a flaw in Claude. It's the strongest reasoning tool available for analyzing data, spotting patterns, and generating strategy. The problem is that without a live data connection, it has no YouTube data to reason about. It fills the gap with plausible-sounding estimates that feel authoritative because Claude's writing is authoritative. The analysis is real. The data underneath it isn't.
TubeLab's YouTube MCP gives Claude the data. If you already use TubeLab's niche finder or outlier finder on the web, the MCP puts that same dataset inside your Claude conversations. No more switching tabs. No more copy-pasting data. Ask Claude and it pulls from TubeLab automatically.
Connect TubeLab to Claude (4 Steps)
No API keys, no terminal, no code. The whole thing is a browser login.
Step 1. Go to claude.ai/settings/connectors and click Add custom connector.
Step 2. Set the name to TubeLab and paste the MCP server URL:
https://public-api.tubelab.net/mcp
Step 3. Click Connect. Sign in to your TubeLab account (or create one) and click Approve.
Step 4. Click Configure on the TubeLab connector and toggle "Always allow" so Claude can call TubeLab's tools without asking permission each time.
Open a new chat and ask anything about YouTube. Claude will call TubeLab's tools automatically.
What Claude Gets Access To
Once connected, Claude has YouTube data tools. It picks the right one based on your question. You just ask in plain language.
Free tools (no credits, no subscription): look up any channel's full profile (subscribers, views, upload cadence), pull any channel's recent videos or Shorts with view counts and engagement data, get full metadata on any video, pull complete transcripts from any video's captions, and read the top comments on any video sorted by engagement.
Paid tools (credits from your TubeLab account): search 400K+ channels with 20+ filters including niche, subscriber range, language, and monetization status (5 credits), find channels similar to any channel you specify (5 credits), search millions of verified outlier videos by topic, views, and performance multiplier (2 credits), and find outliers from channels similar to a given channel (2 credits).
Claude-Only: Pre-Built Research Workflows
This is a feature only Claude has. TubeLab's YouTube MCP server includes prompt workflows built into the protocol, and Claude surfaces them natively in the + menu at the bottom of any chat. No copy-pasting. No typing prompts. Click a workflow and it runs.
- Channel Roast: Drop in any channel URL. Claude pulls data, analyzes performance, and delivers a strategic breakdown of what's working and what isn't.
- Niche Analysis: Specify a category. Claude maps the competitive space, subscriber ranges, and monetization rates.
- Niche Research: Goes deeper. Claude finds outlier videos, cross-references patterns, and surfaces specific content opportunities.
- Trend Report: Identifies topics gaining traction based on recent outlier activity in any niche.
- Video Ideation: Combines outlier data, comments, and transcripts to generate video ideas backed by proven audience demand.
How to Use Claude for YouTube Research (Not Just Lookups)
Pulling a channel's subscriber count is the simple use case. The reason connecting TubeLab to Claude matters is that Claude can chain multiple data pulls into a research workflow and analyze the results. That's where it stops being a lookup tool and starts being a research partner.
Here's how to get the most out of it.
Niche Validation: What to Actually Look For
Most creators validate a niche by browsing YouTube for an hour and deciding based on gut feeling. The problem isn't effort. It's that they don't know which data points actually predict whether a niche will work.
Three things matter when evaluating a niche: how many monetized channels exist in your target subscriber range (competitive density), what view counts look like at that subscriber level (audience demand), and whether channels are growing or stagnant (trajectory).
Here's how to get all three in one Claude conversation:
"Search for monetized YouTube channels in the home renovation niche with 10K-100K subscribers that upload at least weekly in English. How many exist? What's the median view count per video?"
Claude calls TubeLab's channel search and returns actual channels. Count matters: 200+ monetized channels means the niche supports the business model but competition is real. Under 20 means the audience might not be there, or the niche is too small to support monetization.
But count alone isn't enough. Ask Claude to go deeper:
"Pull the last 10 videos from the top 3 channels in those results. Are view counts consistent or do they spike and crash? Are the channels growing or plateaued?"
Consistent views mean a reliable audience. Spike-and-crash patterns mean the channel depends on viral hits and can't sustain watch time between them. This is the difference between a niche with a real audience and a niche that looks active from subscriber counts but doesn't generate steady watch time.
Before doing this in Claude, check TubeLab's niche finder on the web for a quick overview of saturation, CPM estimates, and growth trends. The niche finder tells you if the space is worth entering. Claude tells you exactly what's happening inside it.
Competitor Teardown: The Three-Layer Method
Surface-level competitor analysis looks at subscriber counts and view numbers. It tells you almost nothing useful. The insights come from going three layers deep: performance data, then content structure, then audience response.
Layer 1: Performance patterns. Start with the data.
"Get all the data on @competitor. Pull their last 30 videos. Which 5 performed best and which 5 performed worst? What separates them? Compare titles, durations, topics, and publish days."
Claude gets the channel overview and video list from TubeLab, then analyzes the full set. You might find that their best videos are all 12-15 minutes while their worst are over 20. Or that list-based titles outperform question-based titles. Or that Tuesday uploads consistently beat Friday uploads. These patterns exist in every channel's data. Most creators never see them because finding them manually takes hours.
Layer 2: Script structure. Go deeper into their best performer.
"Get the transcript from their #1 video. How long was the hook before the main content started? How did they structure transitions between sections? Where were the strongest moments in the script?"
Claude pulls the full transcript from TubeLab and analyzes the writing decisions that made the video work. You're not studying what they said. You're studying how they structured what they said: intro length, section pacing, where they placed their strongest points, and how they closed.
Layer 3: Audience response. See how viewers actually reacted.
"Now get the top comments on that video. What are viewers praising most? What questions did they ask that the video didn't answer? What follow-up topics did they request?"
This layer tells you what the audience wanted that they didn't get. The video's biggest gap is your biggest opportunity. The questions viewers asked are content ideas they've already validated by asking.
Three layers, three TubeLab tools, one Claude conversation. Channel data, transcript analysis, comment mining. What used to take an afternoon takes three prompts.
The free TubeLab Chrome Extension adds another layer while you browse YouTube: monetization status, RPM estimates, and performance data visible on any channel with one click.
Content Ideation: Building Ideas from Proof
The strongest video ideas don't come from brainstorming. They come from evidence. Here's a method that uses Claude and TubeLab to build ideas from two data sources: outlier patterns and audience gaps.
Start with outliers. Find what the audience has already proven they want.
"Find outlier videos about personal finance from the last 90 days with at least 10x their channel's average views. Show me the top 10. What topics keep appearing?"
Claude searches TubeLab's outlier database and identifies patterns across the viral videos. If three of the top 10 are about the same subtopic, the audience is hungry for that topic right now.
Then check if the trend is real or a one-off:
"Find outlier videos from channels similar to the top result. Are more creators seeing outsized performance on this subtopic?"
If multiple channels see outlier performance on the same topic, it's a confirmed trend. If it's isolated to one channel, it might not translate to your audience.
For visual browsing of outlier data, TubeLab's outlier finder on the web offers the same dataset in a point-and-click interface.
Then mine comments for gaps. Take the top 3 outlier videos and ask Claude:
"Get the top comments on these 3 videos: [URLs]. What questions appear across multiple comment sections? What did viewers say was missing? What did they disagree with?"
Questions that appear across multiple comment sections are high-confidence content ideas. The audience asked for them more than once, on different videos, from different creators.
Turn it into video concepts:
"Based on the outlier topics and the comment gaps, suggest 5 video ideas. For each, give me a working title, the audience question it answers, and a one-paragraph outline."
Every idea traces back to specific data: an outlier pattern that proved the topic works, and a comment gap that shows the audience wants more. No brainstorming in a vacuum. No hoping a topic lands.
Set Up a Claude Project for Ongoing YouTube Research
One-off research conversations are useful. But the creators who get the most from TubeLab's YouTube MCP treat YouTube research as an ongoing process, not a one-time task.
Claude's Projects feature lets you create a dedicated workspace for YouTube research with a custom system prompt that carries across every conversation in the project. Set one up with instructions like:
"You are my YouTube research assistant. When I ask about a niche, always check competitive density, view count ranges at my target subscriber level, and whether top channels are growing or stagnant. When I ask about a competitor, use the three-layer method: performance data first, then transcript analysis, then comment mining. Always use TubeLab's tools for any YouTube data."
Every conversation you start inside that Project follows those instructions. You don't need to re-explain your research methodology each time. Ask "how's the finance niche looking this month?" and Claude automatically runs the full validation process using TubeLab's tools because the Project instructions tell it to.
This also makes it easy to track changes over time. Run the same niche query monthly. Compare the channel counts, view averages, and outlier patterns. You'll see whether the niche is getting more saturated, whether view counts are rising or falling, and whether the content opportunities from last month are still open or already taken.
Frequently Asked Questions
Does Claude have YouTube data without TubeLab? No. Claude has general knowledge about YouTube but cannot access current channel stats, video performance, or niche data. Any specific numbers about YouTube are from outdated training data or invented. TubeLab gives Claude a live connection to YouTube.
What does it cost? Six tools are free with a free TubeLab account: channel lookups, video details, transcripts, and comments. The four search tools (channels, related channels, outliers, related outliers) consume credits and require a TubeLab subscription at $178.80/year.
Can I use the free tools without a paid subscription? Yes. Create a free TubeLab account, connect it to Claude, and use the six free tools immediately. Look up any channel, pull any video's details, read any transcript, and mine any video's comments at no cost.
What are the pre-built prompt workflows? Research templates built into TubeLab's MCP server that Claude surfaces in the + menu: Channel Roast, Niche Analysis, Niche Research, Trend Report, and Video Ideation. Each runs a multi-step research workflow. You can also browse and copy them from the prompt library.
Can Claude modify my YouTube channel? No. TubeLab's MCP is read-only. Claude can query publicly available data but cannot upload, edit, or change anything on any channel.
Does this work with Claude Projects? Yes. Set up a Project for YouTube research with custom instructions (like "always check niche data before making content recommendations") and Claude will follow those instructions while using TubeLab's tools throughout the conversation.
Sources
TubeLab - YouTube MCP for Claude - Product page and setup overview.
TubeLab - Claude Connector Setup Guide - Step-by-step documentation.
TubeLab - YouTube MCP Server Documentation - Full tool list, authentication, and credits.
TubeLab - Prompt Library - Pre-built research workflows.
Model Context Protocol - Introduction - Official MCP documentation.
Give Claude Real YouTube Data
Claude is the strongest reasoning tool available. It's wasted when the data it reasons about doesn't exist.
Connect TubeLab's YouTube MCP and every question about YouTube gets answered with verified data from actual channels. Six tools are free. Setup takes four steps.
For the full research toolkit (outlier discovery, channel search across 400K+ channels, niche analysis), TubeLab costs $178.80/year. No tiers, no feature gating, no upsells.