Content Fetcher MCP

This MCP server fetches content from various sources including YouTube, RSS feeds, and GitHub releases. It's designed to work with Goose to help track and identify new content.

Features

• YouTube: Fetches videos from the Goose YouTube channel

• RSS Feeds: Fetches blog posts from any RSS feed (including the Goose blog)

• GitHub Releases: Fetches releases from the Block/Goose repository

• Content Tracking: Tracks seen content to identify new items

• Cross-machine Persistence: Stores tracking data in ~/.config/goose/content-fetcher-mcp/

Related MCP server: mcp-morning-brief

Setup

1. Ensure you have Node.js installed (version 14 or higher recommended).

2. Install dependencies:

   npm install

3. Build the project:

   npm run build

Running the MCP Server

This is an MCP server that uses stdio transport and is designed to be registered with Goose.

To start the server directly:

npm start

For development with auto-reload:

npm run dev

Registering with Goose

To register this MCP server with Goose, add it to your Goose configuration. The server uses stdio transport, so it should be configured as a local MCP server in your Goose settings.

Available Tools

1. fetchYoutube

Fetches ALL videos from the Goose YouTube channel.

Parameters: None

Returns: Array of video objects with id, title, url, published_at, and type: "video"

2. fetchRss

Fetches ALL blog posts from any RSS feed.

Parameters:

• url (string): RSS feed URL

Returns: Array of blog post objects with id, title, url, published_at, and type: "blog"

3. fetchGooseBlog

Fetches ALL blog posts from the official Goose blog.

Parameters: None

Returns: Array of blog post objects with id, title, url, published_at, and type: "blog"

4. fetchGithubReleases

Fetches ALL releases from the Block/Goose GitHub repository.

Parameters: None

Returns: Array of release objects with id, title, url, published_at, and type: "release"

5. isNewContent

Checks if a content item has been seen before.

Parameters:

• id (string): Unique identifier for the content

• type (enum): One of "youtube", "blog", or "release"

Returns: { "is_new": true/false }

6. markContentSeen

Marks a content item as seen (typically after posting).

Parameters:

• id (string): Unique identifier for the content

• type (enum): One of "youtube", "blog", or "release"

Returns: { "success": true }

How It Works

1. Fetching: The fetch tools retrieve all available content from their respective sources

2. Filtering: Use isNewContent to check if an item hasn't been seen before

3. Tracking: After processing new content, use markContentSeen to mark it as seen

4. Persistence: Seen content is stored in ~/.config/goose/content-fetcher-mcp/last_seen.json

Example Workflow

// 1. Fetch all YouTube videos
const videos = await fetchYoutube();

// 2. Check which ones are new
for (const video of videos) {
  const result = await isNewContent({ id: video.id, type: "youtube" });
  if (result.is_new) {
    // Process the new video...
    
    // 3. Mark as seen after processing
    await markContentSeen({ id: video.id, type: "youtube" });
  }
}

Configuration

The server is pre-configured with:

• YouTube Channel: Goose channel (UCVLuT_AS687XAJ__-COCRFw)

• Goose Blog RSS: https://block.github.io/goose/blog/rss.xml

• GitHub Repository: block/goose

To customize these, edit the constants in src/server.ts.

Notes

• The server uses stdio transport, making it suitable for local MCP integration with Goose

• Content tracking is persistent across restarts via the last_seen.json file

• All fetch operations return the complete list of content; filtering for "new" items is done separately via isNewContent

Future Improvements

1. Add configuration file support for customizing channels, feeds, and repositories

2. Implement rate limiting and caching to optimize API usage

3. Add more detailed logging and error handling

4. Support for additional content sources

5. Batch operations for checking multiple items at once