MCP Server for javinfo @javinfo/mcp

Install @javinfo/mcp, configure it once, and any MCP host you use, like Claude Desktop, Cursor, or Windsurf, can look up JAV metadata on demand.

What MCP changes is the interface. Instead of prompting the model with “here’s a curl command, trust me,” you configure the server and the model sees javinfo-search and javinfo-movie as things it can just call. It searches when it needs to, fetches a full record when it has a code, and moves on. No header construction, no JSON parsing, no wasted context explaining how HTTP works.

What you get

Two tools, both straightforward:

• javinfo-search. Free-text search by code, title, or actress. Returns a list of matches with IDs.
• javinfo-movie. Fetch one release by exact DVD ID. Optionally pin a provider (r18, javdb, missav, javdatabase) and pass includeImages: true for cover art and samples.

Each result also carries structuredContent with the raw record when the model needs the full metadata.

Setup

{
  "mcpServers": {
    "javinfo": {
      "command": "npx",
      "args": ["-y", "@javinfo/mcp"],
      "env": { "JAVINFO_API_KEY": "jvi_..." }
    }
  }
}

Get your key at javinfo.dev. One line in your config, restart the host, and you’re done.

The JAV metadata space has no official APIs. Every source is a scraper. Wrapping those behind a stable MCP layer means your AI never hits a broken parser, a Cloudflare challenge, or a missing field. It asks for a code and gets structured data back. The MCP server runs on stdio. No ports, no daemon, no Docker.

Source at github.com/javinfo/mcp. Published on npm.