wtf is armenian-national-library-mcp?

This project is an unofficial MCP server that connects AI assistants like Claude Code or Codex to the public catalogue of the National Library of Armenia. MCP stands for Model Context Protocol, a way for AI tools to call external services as structured tools rather than guessing at answers. It is not run or endorsed by the library itself, it simply reads from the library's already-public online repository. Once connected, an assistant can search the library's catalogue for books, periodicals, and manuscripts in Armenian, English, or Russian, filter results by author, subject, year, language, or collection, and look up full metadata for a specific item, including its rights and access status. When a public text version of an item exists, the assistant can read it in bounded chunks, and it can also provide a direct link to an original file on the library's own site so a user never needs the file placed inside the AI's context window. The server is careful about what it claims. Every response includes structured fields showing the data returned, where it came from, any warnings, and whether the result was cut short. It tracks whether an item's rights status is known or unknown, but it never decides on its own whether reusing a file is actually permitted, since public access and legal reuse rights are treated as separate questions. The project is built in TypeScript on Node.js and can run as a local command-line process using standard input and output, or as a stateless web server that other tools connect to remotely. It ships as an npm package, so a user installs it with a single command and adds it to their AI client's configuration, after which the assistant gains a defined set of catalogue-searching tools and stops needing to browse the library's website manually.