youtube-stats-mcp

youtube-stats-mcp

Self-hosted MCP server that lets AI assistants read your own YouTube channel's stats, recent uploads, and upload schedule via the YouTube Data API v3, using read-only OAuth.

Category
Visit Server

README

youtube-stats-mcp

A small, self-hosted MCP server that gives any MCP-compatible AI assistant (Claude, Claude Desktop / Cowork, or any other MCP client) read access to your own YouTube channel's stats and upload schedule.

It runs entirely on your own machine via Docker. Your Google OAuth refresh token stays in a local file that's never committed to git and never leaves your container. Anyone can clone this repo, connect it to their own Google account, and start querying their own channel — no shared credentials, no central server, no data collected by anyone but you.

Features

Five tools, all backed by the YouTube Data API v3:

Tool Description
list_my_channels Profile + stats (subs/views/video count) for the authenticated channel
get_channel_stats(channelId?) Public stats for any channel, or the authenticated one if omitted
get_recent_uploads(maxResults=10) Most recent published videos, each with view/like/comment counts and duration
get_upload_schedule(lookahead=50) Upcoming scheduled/private uploads and premieres/live streams not yet public
get_video_stats(videoId) Views/likes/comments and metadata for a specific video

The server uses the youtube.readonly scope only — it can read your channel's data but cannot upload, edit, delete, or change anything.

Prerequisites

  • Docker and Docker Compose (to run the server)
  • Node.js 18+ (only needed for the one-time OAuth setup script — the server itself runs in Docker)
  • A Google account that owns or manages the YouTube channel you want to track

Quick start

git clone https://github.com/YOUR_GITHUB_USERNAME/youtube-stats-mcp.git
cd youtube-stats-mcp
npm install
  1. Create a Google Cloud OAuth client — one-time setup, see the walkthrough below. You'll end up with a Client ID and Client Secret.

  2. Configure your credentials:

    cp .env.example .env
    

    Open .env and fill in YOUTUBE_CLIENT_ID / YOUTUBE_CLIENT_SECRET from step 1.

  3. Authorize your Google account:

    npm run setup
    

    This prints a Google consent URL. Open it in the browser signed into the YouTube account you want to track, approve access, and the script saves secrets/youtube_token.json for you. That folder is gitignored — it never gets committed.

  4. Build and run the server:

    docker compose up -d --build
    
  5. Confirm it's up:

    curl http://localhost:8787/
    # {"ok":true,"name":"youtube-stats-mcp"}
    
  6. Connect it to your MCP client — see "Connecting it to Claude / Cowork / any MCP client" below.

Create a Google Cloud OAuth client

You need your own OAuth client because this scope (youtube.readonly) is tied to a specific Google Cloud project and can't be shared across users. This takes about five minutes and is free.

  1. Go to the Google Cloud Console and create a new project (or pick an existing one).
  2. APIs & Services > Library — search for YouTube Data API v3 and click Enable.
  3. APIs & Services > OAuth consent screen:
    • User type: External (unless you have a Google Workspace org and want Internal)
    • Fill in an app name, your email as support contact, and your email again as developer contact
    • Scopes: add .../auth/youtube.readonly
    • Test users: add the Google account whose channel you're tracking
    • Save — you can leave this in "Testing" status, but read "Why this keeps happening" below before you do
  4. APIs & Services > Credentials > Create Credentials > OAuth client ID:
    • Application type: Desktop app (important — this is what lets the setup script use an arbitrary local port without pre-registering an exact redirect URI)
    • Name it whatever you like
    • Click Create, then copy the Client ID and Client Secret into your .env file

Connecting it to Claude / Cowork / any MCP client

The server speaks MCP over Streamable HTTP at:

http://localhost:8787/mcp

Add it as a custom/local MCP server in your client's settings — the exact menu depends on the app (look for "Add custom connector" / "Add MCP server"). Once connected, tool names will appear prefixed by however your client names the connection, e.g. mcp__youtube-stats-mcp__list_my_channels.

Multi-channel / brand accounts

The token only grants access to the channel tied to the Google account you authorized in step 3 above. If you (or someone using this repo) manage separate brand-account channels, each one needs its own:

  • OAuth authorization run (npm run setup, signed into that brand account)
  • Token file (change TOKEN_PATH to a different file)
  • Container/port (copy docker-compose.yml, give it a different service name, host port, and TOKEN_PATH)

Re-authorizing (invalid_grant errors)

If a tool call fails with invalid_grant, your refresh token has expired or been revoked. Fix it:

npm run reauth

This reuses the client ID/secret already saved in secrets/youtube_token.json, opens a fresh consent URL, and overwrites the token file. Then restart the server:

docker compose up -d --build

Why this keeps happening (invalid_grant)

If your Google Cloud project's OAuth consent screen is still in Testing publishing status, Google expires refresh tokens for it after 7 days — no exceptions. This is by far the most common cause of recurring invalid_grant errors on a self-hosted setup like this one. Two ways to stop the cycle:

  • Move the consent screen to "In production" (OAuth consent screen > Publish App). For a personal, narrow-scope (youtube.readonly) app this does not require Google's formal verification review in practice — tokens then last until you explicitly revoke them or leave them unused for 6 months.
  • Otherwise, budget for running npm run reauth weekly while in Testing mode.

Other things that produce auth errors, and what they mean:

Error Likely cause
redirect_uri_mismatch Your OAuth client isn't type "Desktop app" — recreate it as that type, or add the exact printed redirect URI to "Authorized redirect URIs" on a Web app client
access_denied / "Access blocked: this app's request is invalid" Your Google account isn't listed as a Test user on the OAuth consent screen (while it's in Testing status)
invalid_grant Refresh token expired (see above) or was revoked at myaccount.google.com/permissions
No refresh_token returned after consent Google only issues a new one on first consent per account+app; remove the app's access at myaccount.google.com/permissions and re-run npm run setup

Security

  • Local-only by default. The server binds to all interfaces on the port you publish, but docker-compose.yml only publishes it to your host's localhost unless you change that mapping yourself. Don't expose port 8787 to the public internet without also setting MCP_AUTH_TOKEN.
  • Optional bearer-token auth. Set MCP_AUTH_TOKEN in .env to require a matching Authorization: Bearer <token> header on every /mcp request — useful if you're reaching the server over Tailscale, ngrok, etc.
  • Read-only scope. youtube.readonly cannot upload, edit, or delete anything on your channel.
  • Credentials never committed. secrets/ and .env are gitignored and excluded from the Docker build context (.dockerignore). Only .env.example (no real values) is tracked.
  • Revoke access anytime at myaccount.google.com/permissions.
  • Runs as a non-root user inside the container (see Dockerfile).

Development (without Docker)

npm install
npm start

Reads the same .env / secrets/youtube_token.json as the Docker setup.

Project structure

youtube-stats-mcp/
├── src/
│   └── index.js        # MCP server + YouTube API calls
├── reauth.js            # OAuth setup / re-auth script (npm run setup | reauth)
├── secrets/              # gitignored — holds youtube_token.json after setup
├── .env.example
├── docker-compose.yml
├── Dockerfile
└── package.json

Contributing

See CONTRIBUTING.md.

License

MIT

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured