API and MCP server

Lire en français

120 682 public-domain books in French, with their exact text chapter by chapter, their source and their licence. Free, no key, no account.

Quick start

To talk to books from your assistant, add the MCP server (below) or simply paste this link into the conversation:

https://liregratuitement.com/en/mcp.md

For code, every route starts with https://liregratuitement.com/api/v1/ and returns JSON:

curl 'https://liregratuitement.com/api/v1/livres?q=boule+de+suif'

OpenAPI specification: /openapi.json. Summary for language models: /llms.txt.

MCP server

The MCP server lets an assistant search for a book, read a chapter and find an exact passage. It generates no text: it serves the published text and its source, and the assistant does the rest. URL: https://liregratuitement.com/mcp

Tool	Parameters	Purpose
`chercher_livres`	requete, auteur, rayon, relu_seulement, limite	Search the catalogue by title, author or keyword.
`fiche_livre`	slug	Book record: verified summary, shelf, difficulty, reading time, source, licence, chapters.
`lire_chapitre`	slug, numero, partie	Exact text of a chapter (split into 30,000-character parts when needed).
`chercher_passage`	slug, texte	Find an exact passage, with its chapter and link.

Install

Claude (claude.ai and desktop app)

Settings, then Connectors, then “Add custom connector”. Name: Lire gratuitement. URL: https://liregratuitement.com/mcp

Claude Code

In a terminal:

claude mcp add --transport http liregratuitement https://liregratuitement.com/mcp

ChatGPT

Settings, then Apps & Connectors, enable developer mode, then “Create” with the server URL: https://liregratuitement.com/mcp

Cursor

In ~/.cursor/mcp.json:

{
  "mcpServers": {
    "liregratuitement": {
      "url": "https://liregratuitement.com/mcp"
    }
  }
}

VS Code (Copilot)

In .vscode/mcp.json:

{
  "servers": {
    "liregratuitement": {
      "type": "http",
      "url": "https://liregratuitement.com/mcp"
    }
  }
}

Any other MCP client

HTTP transport (Streamable HTTP), JSON-RPC 2.0, no authentication: https://liregratuitement.com/mcp

API reference

`GET /livres`

Search the catalogue. Parameters: q, auteur (author), rayon (shelf), relu=1 (reviewed), page, par (per_page, max 50).

curl 'https://liregratuitement.com/api/v1/livres?q=maupassant&relu=1&par=5'

`GET /livres/{slug}`

Book record and chapter list.

curl 'https://liregratuitement.com/api/v1/livres/le-menteur-pierre-corneille'

`GET /livres/{slug}/chapitres/{n}`

Exact chapter text, with previous and next.

curl 'https://liregratuitement.com/api/v1/livres/le-menteur-pierre-corneille/chapitres/7'

`GET /livres/{slug}/texte`

Full plain text, preceded by source and licence.

curl 'https://liregratuitement.com/api/v1/livres/le-menteur-pierre-corneille/texte'

`GET /livres/{slug}/passages?q=`

Find an exact passage: excerpts, chapter number, reading link.

curl 'https://liregratuitement.com/api/v1/livres/le-menteur-pierre-corneille/passages?q=menteur'

Record fields

Field	Meaning
`slug`	stable book identifier
`titre, auteur`	display title and author
`rayon, difficulte`	shelf and difficulty (easy, classic, demanding)
`quarts_d_heure, mots`	reading time in quarter-hours (180 words per minute) and word count
`relu, qualite_ocr`	proofread text, or digitised with its share of recognised words
`resume, resume_verifie`	summary, published only when checked against the text
`source, licence, usage_commercial`	provenance, licence, and false for Gallica texts
`fiche`	book page URL
`texte`	exact text of the source edition, paragraphs separated by a blank line; italics stay marked with underscores, _like this_, as in Markdown

Use cases

Talk to a book

Install the MCP server, or paste a book's ia.en.md link into your assistant: it reads the requested chapter, explains a passage, asks comprehension questions, and never invents a quotation.

Teaching

Prepare a class from the exact text of a chapter, find a quotation for a commentary, pick short and easy books by reading time and difficulty.

Checking a quotation

The passages route confirms that a sentence is in the text, with its chapter and link.

Building a reading app

Catalogue, records, chapters: enough for an e-reader, a book club or a daily serial.

Accessibility and text-to-speech

Plain text suits reading aloud and display settings.

Research and digital humanities

Full text and consistent metadata (source, licence, text quality) for your corpora, within licence terms.

Grounding an AI in its sources

Answers based on the published text, with the reference: that is what the MCP server is for. No bulk export here: for model training, see the dedicated open corpora.

Limits and errors

20,000 requests per day per IP address. Public responses cached for one hour. All routes accept browser calls (open CORS).

Code	Meaning
404	book, chapter or route not found
429	daily quota exceeded, try again tomorrow
503	temporary incident, retry in a minute

Licences

The works are in the public domain, and every response carries the text's source and licence. Wikisource transcriptions: CC BY-SA 4.0 (attribute, share alike). Bibebook: CC BY-SA 3.0. Gallica-sourced texts: commercial reuse requires a licence from the Bibliothèque nationale de France, flagged as usage_commercial: false. Our metadata (shelves, difficulty, reading time, summaries): CC BY 4.0, with a link to the book page.

Questions, larger needs, an error in a text: contact@liregratuitement.com