tdarr-mcp

A Model Context Protocol (MCP) server for Tdarr — the distributed media transcoding/health-check automation platform. It gives AI assistants (Claude Desktop, Claude Code, and any MCP-compatible client) programmatic control over a Tdarr server through its HTTP API.

Speaks MCP over stdio and exposes 105 tools across 12 categories.

Features

All tools are prefixed tdarr_* (e.g. tdarr_get_nodes, tdarr_search_db, tdarr_run_automation).

Prerequisites

• Node.js ≥ 18 (uses the native fetch API), or Bun.

• A running Tdarr server reachable over HTTP.

Installation & build

git clone https://github.com/maximeallanic/tdarr-mcp.git
cd tdarr-mcp

# with Bun
bun install
bun run build      # tsc → dist/

# …or with npm
npm install
npm run build

This compiles src/ to dist/.

Configuration

The server reads its target from environment variables:

Usage with an MCP client

Add the server to your MCP client configuration (e.g. Claude Desktop / Claude Code mcpServers block):

{
  "mcpServers": {
    "tdarr": {
      "command": "node",
      "args": ["/absolute/path/to/tdarr-mcp/dist/main.js"],
      "env": {
        "TDARR_URL": "http://your-tdarr-host:8265"
      }
    }
  }
}

During development you can run the TypeScript entry directly with Bun (no build step):

TDARR_URL=http://your-tdarr-host:8265 bun run src/main.ts

How it works

Each tool is a thin declarative mapping (TdarrToolDef) to a Tdarr API endpoint: HTTP method, path (with :param substitution), and an input JSON schema. POST bodies are wrapped in { data: ... } by default (Tdarr's convention) unless a tool sets dataWrapped: false. Responses are returned as pretty-printed JSON text.

License

MIT © Maxime Allanic