Getting Started with Your AI Tool

With Watt running and extensions installed, connect your AI agent using the instructions for your tool below.

Claude Desktop

Claude Desktop only supports stdio, so we use supergateway to bridge it to Watt’s SSE endpoint. Supergateway runs as a child process that Claude launches automatically. Nothing extra to install or start manually.

1

Open the config file

In Claude Desktop, go to Settings → Developer → Edit Config.

This opens ~/Library/Application Support/Claude/claude_desktop_config.json.

2

Add the mcpServers block

Add or merge the following into the config file:

{
  "mcpServers": {
    "watt": {
      "command": "npx",
      "args": ["-y", "supergateway", "--sse", "http://localhost:9000/mcp/global/sse"]
    }
  }
}

If you already have other MCP servers configured, add "watt": { ... } inside the existing mcpServers object. Don’t replace the whole file.

3

Restart Claude Desktop

Save the file and fully quit and relaunch Claude Desktop. It will start supergateway automatically and discover all installed Watt tools.

Claude (web)

Claude on the web connects to MCP servers from the cloud, so it can’t reach localhost directly. We use ngrok to create a secure HTTPS tunnel from the internet to your Mac.

1

Install and start ngrok

If you don’t have ngrok, install it:

brew install ngrok

Then start the tunnel:

ngrok http 9000

Copy the HTTPS URL shown in the terminal, something like https://abc123.ngrok.app.

2

Add the connector in Claude

In Claude web, go to Settings → Connectors → Add custom connector.

Fill in the fields:

  • Name: Watt
  • URL: your ngrok URL with the MCP path appended:
https://abc123.ngrok.app/mcp/global/sse

Save the connector.

3

Keep ngrok running while you use Watt

ngrok must stay running for the connection to work. The tunnel URL changes each time you restart ngrok unless you have a paid ngrok account with a static domain. Update the integration URL if it changes.

ChatGPT

ChatGPT connects to MCP servers from the cloud, so it can’t reach localhost directly. We use ngrok to create a secure HTTPS tunnel from the internet to your Mac.

Requires ChatGPT Pro, Team, Enterprise, or Edu.

1

Install and start ngrok

If you don’t have ngrok, install it:

brew install ngrok

Then start the tunnel:

ngrok http 9000

Copy the HTTPS URL shown in the terminal, something like https://abc123.ngrok.app.

2

Enable Developer Mode in ChatGPT

In ChatGPT, go to Settings → Apps → Advanced Settings and toggle Developer Mode on.

3

Create a new app

A New App form will appear. Fill in the fields:

  • Name: Watt
  • Description: YouTube analytics and research tools
  • MCP Server URL: your ngrok URL with the MCP path appended:
https://abc123.ngrok.app/mcp/global/sse

Save the app.

4

Use Watt in a conversation

When chatting, click the + button at the bottom of the input field, select More, then choose Watt from the list. Watt’s tools are now available for that conversation.

5

Keep ngrok running while you use Watt

ngrok must stay running for the connection to work. Update the MCP Server URL in your app settings if the tunnel URL changes.

OpenClaw

OpenClaw uses mcporter as its MCP bridge. Install it once, configure it, and OpenClaw picks it up automatically.

1

Install mcporter

npm install -g mcporter
2

Create the config directory and file

mkdir -p ~/.mcporter

Then create ~/.mcporter/mcporter.json:

{
  "mcpServers": {
    "watt": {
      "url": "http://localhost:9000/mcp/global/sse"
    }
  }
}
3

Verify the connection

Run mcporter list in your terminal. You should see Watt’s tools listed.

4

Connect OpenClaw to Watt

Pick either approach:

Option A: System prompt

Add the following to your OpenClaw system prompt (or say it at the start of each conversation):

Always use mcporter to interact with Watt for any YouTube analytics, research, or transcription tasks.

Option B: Generate a dedicated CLI

Use mcporter to generate a named CLI for Watt:

mcporter generate watt

Then instruct OpenClaw (via system prompt or at conversation start):

Use the watt CLI to handle all YouTube analytics, research, and transcription tasks.

OpenClaw will invoke the watt command directly, keeping Watt usage explicit and consistent.

What to ask

Once connected, here are prompts to get started:

Channel performance

  • “How did my last 10 videos perform compared to my channel average?”
  • “Which of my videos got the most watch time this month?”
  • “Show me my top traffic sources for the past 30 days”

Competitor research

  • “Add [channel name] to my competitor watch list and summarise their last 5 videos”
  • “Which competitor videos in my niche got the most engagement this week?”

Transcription

  • “Transcribe this video: [YouTube URL]”
  • “Summarise the key points from this interview: [YouTube URL]”
  • “Find every timestamp where [topic] is mentioned in this video: [YouTube URL]“

Creator strategy

  • “What does the YouTube algorithm prioritise for channels my size?”
  • “How should I structure my video hooks for better retention?”
  • “What’s a good thumbnail strategy for my niche?”

Tips

Ask follow-up questions. Your agent keeps conversation context, so “now compare that to last month” or “why do you think that happened?” will use data already fetched.

Combine capabilities. Pull a transcript and analyse its structure in the same conversation. Fetch competitor stats then ask for a positioning recommendation.

Extensions start on demand. Watt auto-starts any extension the first time a tool is called. No need to activate them manually before asking.