BestReads MCP Server 📚🤔

This is a remote MCP server, built on Cloudflare Workers, that provides personalized book recommendation.

This was built using Cloudflare's guide on deploying remote MCP servers. It uses the Agents SDK to build the MCP server, Durable Objects to persist the user's book preferences, Workers AI to generate book recommendations, and Cloudflare's OAuth Provider library to add GitHub as an authentication provider. The MCP server supports Server-Sent Events (/sse) and Streamable HTTP (/mcp) transport methods.

Get Started

To try it out, connect to https://bestreads.dinas.workers.dev/sse or https://bestreads.dinas.workers.dev/mcp, if your MCP client supports Streamable HTTP. Or, deploy it yourself using the Deploy to Cloudflare button + instructions below.

Available Tools

• getProfile - View your reading history and preferences

• addGenre - Add favorite book genres

• addFavoriteAuthor - Add authors you enjoy

• addBookRead - Track books you've read

• addDislikedBook - Mark books you didn't enjoy

• addDislikedAuthor - Authors to avoid in recommendations

• clearPreferences - Reset all preferences

• getBookRecommendations - Get AI-powered personalized book suggestions

Deploy the MCP server

Setup

1. Clone the repository

  git clone <your-repo-url>
  cd bestreads-mcp-server
  npm install

2. Create a GitHub OAuth App

• Once you create teh OAuth App, set the Authorization callback URL to https://your-worker-domain.workers.dev/callback

• Note the ClientID and Client Secret. You will add those to your Wrangler file.

• (Optional) Generate Cookie Encryption Key

3. Upgrade your wrangler.toml file

[vars]
GITHUB_CLIENT_ID = "your_github_client_id"
GITHUB_CLIENT_SECRET = "your_github_client_secret"
COOKIE_ENCRYPTION_KEY = "your_32_byte_hex_key"

[[kv_namespaces]]
binding = "OAUTH_KV"
id = "your_kv_namespace_id"

[[durable_objects.bindings]]
name = "MCP_OBJECT"
class_name = "MyMCP"

[[durable_objects.bindings]]
name = "USER_BOOK_PREFERENCES"
class_name = "UserBookPreferences"

4. Deploy to Cloudflare Workers wrangler deploy