Model Context Protocol (FastMCP) Server

TubeSentry provides native MCP support, enabling AI models in Claude Desktop, Claude Code, and Cursor to audit ad placements and inspect YouTube channels natively.

---

Configuration

Claude Desktop

Edit your claude_desktop_config.json:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "tubesentry": {
      "command": "uvx",
      "args": ["--from", "tubesentry", "tubesentry-mcp"]
    }
  }
}

Cursor IDE

Create or update .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "tubesentry": {
      "command": "tubesentry-mcp"
    }
  }
}

---

Exposed MCP Tools

1. audit_placement_report

Audits a Google Ads or DV360 placement CSV export for waste categories.

• Arguments: csv_path (required), target_languages, target_cpa, ad_url, ad_text, output_dir.
• Output: Full audit summary including waste breakdown, quality score, top offenders, and creative alignment metrics.

2. inspect_youtube_channels

Looks up specific YouTube Channel IDs or URLs in the TubeSentry Intelligence Database.

• Arguments: channel_ids_or_urls (list of strings).
• Output: Channel title, waste category, confidence, and audit rationale.

3. profile_creative_intent

Profiles a YouTube ad URL or brief with Gemini.

• Arguments: ad_url, ad_text.
• Output: Target persona, recommended contextual topic entities, and negative anti-personas.

---

Example Natural Language Workflows

User: "Audit ~/Downloads/pmax_october.csv for Made-for-Kids toddler channels and export the exclusion list to ~/Desktop/exclusions.csv"

Claude: Calling tubesentry.audit_placement_report(csv_path="~/Downloads/pmax_october.csv", output_dir="~/Desktop")
...
Audit complete! 38.4% of your spend (€14,200) was wasted on toddler and passive music channels. 
Exclusion file saved to ~/Desktop/google_ads_reactive_exclusions.csv.