Trakt

What is Trakt?

Trakt is an MCP (Model Context Protocol) server that bridges AI language models with the Trakt.tv API, allowing LLMs to access real-time entertainment data and personal Trakt viewing history. It is built with a domain-focused architecture using FastMCP, providing clean separation across authentication, shows, seasons, episodes, movies, people, user data, comments, search, and check-in functionality.

How to use Trakt?

Trakt can be run via Docker, uvx (without cloning), or local installation (requires Python 3.12+). You need Trakt API credentials (client ID and secret). Configure it in Claude Desktop or MCP hub by adding the server to your MCP configuration file with the appropriate command and environment variables.

Key features of Trakt

• Access trending, popular, anticipated, and top-grossing content
• View personal watched shows, movies, and progress
• Get personalized movie and show recommendations
• Manage ratings, watchlist, playback progress, and history
• Browse comments, reviews, and cast/crew for any content
• Secure OAuth authentication with device code flow

Use cases of Trakt

• Ask an AI assistant for currently trending movies or shows
• Get personalized recommendations based on your Trakt history
• Manage your watchlist or check in to shows via natural language
• Look up cast, crew, ratings, and reviews for any movie or show
• Track your viewing progress and see what to watch next

FAQ from Trakt

Do I need a Trakt account to use this server?

Public Trakt data (trending, popular, box office, etc.) does not require authentication. Personal features like watch history, ratings, recommendations, and check-ins require a Trakt account and OAuth authentication via device code flow.

How do I authenticate with Trakt?

Authentication uses the Trakt device code flow. Your OAuth token is persisted to ~/.trakt-mcp/auth_token.json (or a custom path via TRAKT_AUTH_TOKEN_PATH), so authorization survives across uvx invocations.

What are the runtime and dependency requirements?

Running locally requires Python 3.12 or newer. Alternatively, you can use Docker or run with uvx (requires uv installed). The server is configured in Claude Desktop as a command-line MCP tool.

How is pagination handled for list endpoints?

You can pass a page parameter for single-page results with pagination metadata, or omit it to auto-paginate up to a given limit. Use limit=0 to fetch all available results (capped at 100 for safety).