LINKEDMASH_TOKEN=your_linkedmash_api_token_here
LINKEDMASH_WEBAPP_URL=https://www.linkedmash.com  # URL to the webapp for setup endpoints
NODE_ENV=development

Installation

# Install dependencies
npm install

# Run the service
npm start

# For development with auto-restart
npm run dev

Docker Setup

# Build and run with docker-compose
docker-compose up mcp-service

# For development mode
docker-compose --profile dev up mcp-service-dev

Available Tools

1. `list_bookmarks`

Lists user's LinkedIn bookmarks with optional filtering and pagination.

Parameters:

limit (number, 1-100): Number of bookmarks to retrieve
cursor (string): Pagination cursor for next page
is_unread_only (boolean): Filter to only unread bookmarks
hide_archived (boolean): Hide archived bookmarks
media_type (enum): Filter by media type (thread, media, replies, notes, links)
author (string): Filter by author username
tag (string): Filter by tag
posted_from/posted_to (string): Filter by post posting date range
bookmarked_from/bookmarked_to (string): Filter by bookmark date range
sort_by (enum): Sort order for results

2. `search_bookmarks`

Searches through bookmarks using a text query.

Parameters:

q (string, required): Search query
All the same filtering parameters as list_bookmarks

3. `get_bookmark_count`

Gets the total count of bookmarks with optional filtering.

Parameters:

All filtering parameters except limit, cursor, and sort_by

4. `setup_vector_store`

Initializes the vector store for semantic search. Call this when vector/semantic search fails with "Vector Store Not Setup" error.

Parameters: None

Note: This is an async operation that may take a few minutes to complete for large bookmark collections.

API Integration

The service integrates with the Linkedmash API at https://api.linkedmash.com/v1/bookmarks using Bearer token authentication.

Error Handling

The service includes comprehensive error handling and validation:

Invalid API tokens return authentication errors
Malformed requests are validated using Zod schemas
API errors are properly formatted and returned to the client

Development

The service is built using:

@modelcontextprotocol/sdk for MCP protocol implementation
Zod for request/response validation
Standard Node.js fetch for API calls

Usage with AI Assistants

This MCP service can be integrated with AI assistants that support the Model Context Protocol, allowing them to access and manage Linkedmash bookmarks on behalf of users.