LinkSync AI

Your links, synced smarter by MCP + AI.

LinkSync AI is a bookmark manager + MCP bridge that lets AI clients work with your saved links using natural language.

Live Demo

• https://ai.shivaprogramming.com
• 

SEO Key Points

• Target keywords: bookmark manager, MCP integration, AI bookmark sync, link management
• Unique value proposition: LinkSync AI combines MCP + AI integration so assistants can work directly with bookmark context.
• Call-to-action for indexing: Login with GitHub or Google
• Structured data: SoftwareApplication schema markup is included in the dashboard entry page.

What Problem This Solves

Most bookmark tools are built for manual use in the browser. AI clients (Cursor, Claude Desktop, VS Code MCP, etc.) cannot easily access those bookmarks in a structured way.

This project solves that by providing:

• A web dashboard to manage bookmarks (create, edit, favorite, search)
• A backend API for bookmark operations
• An MCP server so AI clients can read and use bookmark data through MCP tools

Technology Used

• TypeScript for API and MCP server code
• Node.js + Express for HTTP routes and dashboard serving
• PostgreSQL for bookmark persistence
• OAuth login for user sessions in the dashboard
• Model Context Protocol (MCP) with stdio transport for client integrations
• HTML/CSS/Vanilla JS for a lightweight dashboard UI
• Docker Compose for local or server containerized runtime

Project Structure

• server/api.ts - REST API, dashboard routes, session user endpoints
• server/mcp/index.ts - MCP server entry point (stdio)
• dashboard/index.html - Main dashboard UI (Bookmarks, Health, Account, MCP Setup)
• dashboard/mcp-setup.html - Guided MCP setup screen
• configs/mcp/ - MCP client configuration examples
• docs/mcp-clients.md - Additional client setup notes

Run Locally

npm install
npm run build

Start servers:

• API server: npm run start:api
• MCP server: npm run start:mcp
• API dev mode: npm run dev:api
• MCP dev mode: npm run dev:mcp

Local Mode Without Login (read existing DB bookmarks)

Use this when you want to see all bookmarks from your local database before enabling OAuth.

PowerShell (current terminal session):

$env:ALLOW_LOCAL_FALLBACK="true"
$env:SESSION_COOKIE_SECURE="false"
npm run dev:api

Then open:

• http://localhost:3001/bookmarks

Notes:

• With ALLOW_LOCAL_FALLBACK=true, /api/bookmarks reads from the local default user and does not require login.
• Existing bookmarks already in PostgreSQL will be listed immediately.
• For Docker Compose, you can set ALLOW_LOCAL_FALLBACK=true in your environment before docker compose up.

MCP Client Integration

Recommended flow (UI based)

1. Open the dashboard and log in.
2. Go to the MCP Setup tab (or /mcp-setup).
3. Click Generate Token.
4. Use Copy Token and Copy Config buttons.
5. Paste the config into your MCP client settings.

MCP config template

Use this as the base template in your MCP client:

{
	"mcpServers": {
		"bookmark": {
			"command": "npx",
			"args": ["-y", "github:shivasharma/website_bookmark_mcp"],
			"env": {
				"BOOKMARK_API_BASE_URL": "https://ai.shivaprogramming.com",
				"BOOKMARK_API_TOKEN": "paste-token-here"
			}
		}
	}
}

MCP Usage Examples

After client setup, you can ask your AI assistant things like:

• "Find my bookmarks about Docker health checks"
• "Show my favorite MCP resources"
• "List recent links tagged with TypeScript"

Dashboard Features

• Bookmarks tab with favorites and full list views
• System Health mini dashboard
• About Project with README view
• Account tab with session state
• MCP Setup tab with token generation and one-click copy actions