mark-coach-mcp

Local MCP server that turns Mark Builds Brands' YouTube knowledge into an AI coaching assistant for ecommerce and Facebook Ads.

Ask Claude to analyze your Ads Manager screenshots, debug campaigns, or get strategic advice — and it'll respond using Mark's actual frameworks, vocabulary, and mental models from his 100+ videos.

How it works

1. Downloads transcripts from Mark's YouTube channel

2. Indexes them locally into a vector database (ChromaDB)

3. Exposes a search_mark_knowledge tool via MCP

4. A Claude skill activates the persona and queries the knowledge base

All data stays on your machine. No API keys required.

Related MCP server: noapi-google-search-mcp

Quick start

git clone https://github.com/BlueNacho/mark-coach-mcp
cd mark-coach-mcp
chmod +x setup.sh
./setup.sh

The setup script will:

• Install uv (Python package manager) if not present

• Install yt-dlp if not present

• Install Python dependencies

• Download transcripts from Mark's channel (or use your own)

• Index everything into a local vector DB

• Print the exact command to connect it to Claude

Connect to Claude Code

After running setup, add the MCP server:

claude mcp add mark-coach -- uv --directory "/path/to/mark-coach-mcp" run src/server.py

Or add manually to ~/.claude.json:

{
  "mcpServers": {
    "mark-coach": {
      "command": "uv",
      "args": ["run", "src/server.py"],
      "cwd": "/path/to/mark-coach-mcp"
    }
  }
}

Adding new videos

When Mark publishes new content, just run:

./setup.sh

It skips already-indexed videos and only processes new ones.

Using your own transcripts

If you already have .vtt files, copy them to the transcripts/ folder and run:

uv run src/indexer.py transcripts/

Using the skill in Claude Code

Copy skills/mark-coach.md to your Claude skills directory, then activate with /mark-coach in any conversation.

Project structure

mark-coach-mcp/
  setup.sh          ← one-command setup
  pyproject.toml    ← Python dependencies (managed by uv)
  src/
    indexer.py      ← converts .vtt transcripts → ChromaDB
    server.py       ← MCP server with search tool
  skills/
    mark-coach.md   ← Claude skill / persona definition
  transcripts/      ← your .vtt files (gitignored)
  data/             ← ChromaDB vector store (gitignored)

Requirements

• macOS or Linux

• Python 3.11+

• uv (installed automatically by setup.sh)

• Claude Code or Claude Desktop