Triangle pointing up
MCP Server

YouTube MCP Server

Access YouTube data via Claude, ChatGPT, Codex, Cursor, and any MCP-compatible AI agent.

TubeLab runs a YouTube MCP server that lets Claude, ChatGPT, Codex, Cursor, and any other MCP-compatible agent read YouTube data directly from chat — channel stats, viral outliers, transcripts, comments, and more.

No code, no API wiring, no scripts.

YouTube MCP server URL:

https://public-api.tubelab.net/mcp

Access YouTube data from your favourite AI

Pick your AI client and follow the setup guide:

Authentication

The YouTube MCP server supports two authentication methods depending on the client:

  • OAuth — used by Claude and ChatGPT. Log into TubeLab in a browser redirect, approve access, and the connector handles session refresh automatically. No API key to manage.
  • API key (Bearer token) — used by Codex, Cursor, and any custom agent. Create a key and paste it as the Bearer token. Treat it like a password.

Available YouTube data

Once connected, your AI agent has access to 11 tools covering the full TubeLab YouTube dataset:

ToolWhat it doesCost
search_channelsSearch YouTube channels by keyword with filters (subscribers, views, language, monetization, etc.)5 credits
search_related_channelsFind channels similar to a given channel5 credits
search_outliersFind viral or underperforming videos across YouTube by keyword2 credits
search_related_outliersFind outlier videos from channels similar to a given channel2 credits
get_channelGet info about a YouTube channel (accepts ID, URL, or @handle)free
get_channel_videosGet recent videos from a channelfree
get_channel_shortsGet recent Shorts from a channelfree
get_videoGet detailed info about a specific videofree
get_video_transcriptGet the transcript/captions of a videofree
get_video_commentsGet top comments from a videofree
get_credits_balanceCheck your TubeLab credit balancefree

You don't need to call these by name. Just ask your agent what you need in plain language and it picks the right tools automatically.

Prompts

Ready-to-use prompt templates for channel teardowns, niche research, trend reports and more — paste them into any MCP-connected chat.

What you can ask

Once the YouTube MCP server is connected, interact with YouTube data in plain language — niche research, competitor analysis, outlier discovery, transcript extraction, comments mining, and more. For battle-tested prompts you can copy-paste straight into chat, head to the prompts library.

Agents without MCP support

For agents that can't use MCP (custom scripts, curl-based automation, older chat interfaces), we also publish a skill file documenting every endpoint with curl examples. Include it in your agent's system prompt or context and it will generate correct requests against TubeLab's YouTube data endpoints.

curl -o SKILL.md https://tubelab.net/docs/skill.md

The skill is auto-generated from the OpenAPI spec and stays in sync with the API.