# NDL Japan Library Search AI Agent Connect

> Japan's National Diet Library as an MCP: search 100M+ records — books, journals, dissertations, digitized Meiji/Edo classics — by title, author, publisher or ISBN, with NDL viewer links. Keyless.

## Overview
- **Category:** government-public-data
- **Price:** Free
- **Endpoint:** https://edge.vinkius.com/vk_preview_1KHZON53bo83KKBQFQ5L9hXvIX9xS0yeWF03PVow/ai-agent-connect
- **Tags:** japan, library, bibliographic-data, isbn-lookup, digitized-books, research

## Description

The **National Diet Library (NDL)** — Japan's national library and one of the largest collections on Earth — as a single MCP server.

### What you can do
- **Search 100M+ records** — every book, journal article, dissertation, map and government document catalogued in Japan, matched by title, author, publisher or any field
- **ISBN lookup** — authoritative Japanese bibliographic data for any book: exact kanji title, author, publisher, dates, volumes/pages, subjects
- **Digitized treasures** — the NDL Digital Collections hold free scans of Meiji and Edo era books, ukiyo-e, historical documents and pre-war journals

### What lives here
The memory of Japan: 400 years of printing, every PhD thesis since Japan had universities, the complete congressional library of the Diet, and a digitization program that made centuries of rare Japanese books free to read online.

### Who is this for?
Researchers, translators, historians, bibliographers, librarians, book collectors and AI agents that need authoritative Japanese bibliographic data or want to locate digitized Japanese primary sources.

## Tools

### find_book_by_isbn
The fastest way to get authoritative Japanese bibliographic data for a book. NOTE: the NDL API requires the hyphenated ISBN form — this tool inserts the hyphens automatically.

Look up one book in the NDL catalog by ISBN-13 or ISBN-10

### search_national_library
Query modes: title= (book/article titles), any= (all fields), creator= (author), publisher=. Japanese terms work best; English also matches many records. Returns title, author, publisher, date, ISBN, subjects, categories and NDL page URL. NOTE: the API has no server-side pagination — set cnt to the window you need (max ~200).

Search the National Diet Library of Japan: 100M+ books, journals, dissertations, digitized classics and government documents

### search_digitized_history
Ideal for 明治/江戸 era books, historical documents, old maps and journals that are digitized and free to read. Returns digitized item metadata with NDL viewer links.

Search NDL Digitized Collections: scanned Japanese classics, Meiji-era books, rare documents (all free)

## Prompt Examples

**Prompt:** 
```
Find books about ramen in Japan's national library
```

**Response:** 
```
📚 **NDL search: ラーメン (title)**

10,983 records. Top hits:
- 嗚呼!このラーメンがうまい : 首都圏版 — guide to the Kanto's best bowls
- 嗚呼!このラーメンがうまい : 関西版 — Kansai edition

From 1970s food culture to modern academic studies of ramen globalization. Want a filtered look at the academic titles?
```

**Prompt:** 
```
Look up ISBN 978-4-12-390001-1 in the NDL catalog
```

**Response:** 
```
📖 **ISBN lookup via NDL**

Title: 数え方の辞典 (Dictionary of Counting Words)
Author: 飯田朝子 · Publisher: 朝日新聞出版

Japanese counters (本/枚/匹...) are famously hard — this is the standard reference. I can pull related counter-dictionary titles too.
```

**Prompt:** 
```
Are there digitized Meiji-era books about the railway?
```

**Response:** 
```
🎴 **NDL Digitized Collections: 明治 鉄道**

Yes! Free scans include:
- Meiji railway construction reports (official documents)
- Early travel guides with train timetables from 1889
- Illustrated books on the Tokaido line

All readable in the NDL viewer. I can search a specific year or region next.
```

## Frequently Asked Questions

**Do I need an API key or registration?**
No. The NDL Search OpenSearch API is fully public — all tools work without authentication.

**Does the API support pagination?**
The NDL OpenSearch endpoint has no server-side pagination (tested: pagenum/first/start params are ignored). Each call returns the top 'cnt' results (up to ~200). For deeper results, refine the query terms instead of paging.