library-mcp

The ForgeMesh Library MCP server is a product of ForgeMesh Labs, positioned as a paid public-domain knowledge base for AI agents. It provides full-text search, book metadata, chapters, quotation excerpts, a flagship "ask the book" retrieval question-answering feature, and more than 100 curated topical bookshelves, covering a growing corpus of public-domain books. No license, API key, or account is required; payment is per call in USDC via the x402 protocol on the Base chain.

Purpose: This service is built specifically for education-safe agent use cases, including tutoring agents that need real quotations rather than paraphrases, reading apps and book club bots, quiz and lesson generators that cite public-domain source texts, and any agent that needs a large, inexpensive, permissionless text corpus without the rate limits of free APIs.

Key capabilities:

• Full-text search: retrieve content from the indexed text.
• Book metadata, chapters, quotation excerpts.
• Flagship "ask the book" (ask_book) retrieval question-answering, covering works such as Alice's Adventures in Wonderland, The Art of War, The Count of Monte Cristo, Crime and Punishment, Don Quixote, Dracula, Frankenstein, Great Expectations, Grimms' Fairy Tales, Jane Eyre, Meditations, Moby-Dick, The Picture of Dorian Gray, Pride and Prejudice, Sherlock Holmes, The Iliad, The Odyssey, The Republic, The Time Machine, The Adventures of Tom Sawyer, Treasure Island, Walden, War and Peace, Wuthering Heights, and others. Other books can be accessed via search_literature or search_books combined with get_chapter / get_quotes.
• Bookshelf browsing (browse_shelf): 10 hand-picked categories (Adventure, Children's Classics, Classic Literature, Economics and Politics, Gothic Horror, History, Philosophy, Poetry, Science and Nature, Shakespeare and Drama), plus more than 100 topical bookshelves automatically generated from real Gutenberg subject/bookshelf metadata (Gothic fiction, Russian literature, detective fiction, banned books lists, etc.). The complete bookshelf catalog can be browsed for free, with no payment or tool call required:https://library.forgemesh.io/llms.txt and https://library.forgemesh.io/openapi.json。

How to use:

• Quick start (Claude Desktop / Claude Code / any MCP client): requires Node.js 20+ and a dedicated low-balance Base wallet funded with USDC. WALLET_PRIVATE_KEY is a dedicated low-balance Base wallet for x402 micropayments; never use your main wallet. $1 of USDC can buy dozens to hundreds of calls depending on the route.
• The server refuses to sign for any payee other than the Library wallet, any network other than Base mainnet, any asset other than USDC, or any amount above the caps ($0.02 per call, $10 per session). The environment variables X402_MAX_PRICE_USD and X402_SESSION_BUDGET_USD can only lower these caps. Requests are same-origin, with a 60-second timeout, a 2 MB limit, and no redirect following.
• Prices are set by the live service and may change—the actual cost always comes from the x402 402 challenge at call time, not from a fixed table.
• Payment method: no registration, API key, or subscription required. The first request to any tool returns an HTTP 402 challenge; this MCP server signs a USDC payment authorization (EIP-3009) and automatically retries. The result is returned in the same response, and on-chain settlement details are under _payment when available.
• Prefer raw HTTP? The complete agent-readable interface:https://library.forgemesh.io/llms.txt（面向智能体的一页摘要）、https://library.forgemesh.io/openapi.json（OpenAPI 3.1, including x402 payment metadata, complete input schemas, and examples for each route),https://library.forgemesh.io/.well-known/x402.json（x402 Discover the list).

Use cases: tutoring agents, reading apps, book club bots, quiz and course generators, and any agent that needs a large, cheap, license-free text corpus without rate-limited free APIs.

Limitations:

• Retrieval only: every answer is a real, cited passage pulled from the indexed text. Nothing is LLM-generated or summarized, so it is suitable for citation.
• The corpus is actively growing—starting from hundreds of books, moving from 17,000+ toward about 62,000. Do not hardcode the number of books downstream; ask search_books or search_literature and let the live index answer.
• The entire corpus is in the public domain, partly from the Project Gutenberg collection. Not affiliated with or endorsed by Project Gutenberg.
• Requires Node.js 20+ and a dedicated low-balance Base wallet.
• Payments use USDC on Base mainnet (eip155:8453).
• No account or API key required; x402 payment is the only credential.
• Can be used for tutoring or course content—this is the intended use. Retrieval-only, cited-passage answers are better for educational safety than generated summaries.

Built by ForgeMesh Labs · Powered by the x402 protocol · MIT License.