Overview

Project Gutenberg's library of 78,000+ public-domain books. Search by title, author, topic, or language, fetch full book metadata, and read supported UTF-8 plain-text and HTML editions in manageable chunks.

Tools

The published gutenberg-guide skill covers deciding whether a book can be read, ruling out search blind spots before calling a book missing, finding openings and chapters past front matter, paging long texts, and ranking editions. Plugin installs include it; MCP clients can discover it through the Skills extension and its skill resource.

Capability reference

gutenberg_search_books _tool

• Filters: query (title/author words), topic, translator, languages (catalog codes exactly as results report them, two or three letters: en, not eng), author_year_start/author_year_end, ids, mime_type, and copyright ("true", "false", "null"; false means public domain in the USA). Live fallback rejects queries over four words and the translator filter
• With both author-year bounds, one author must meet both. The live fallback checks the bounds separately and says so in a notice
• sort is popular (default) or ascending/descending by Gutenberg ID. Up to 32 books per page; use totalCount and hasMore to page
• Each result carries is_readable (readable by gutenberg_get_text) and has_plain_text. Result rows omit translators; gutenberg_get_book reports them. Fails with translator_requires_mirror, query_requires_mirror, page_out_of_range, or catalog_unavailable

gutenberg_get_book _tool

• Returns the full formats map (MIME type → download URL) and authors, translators, and editors with birth/death years
• Also carries is_readable, has_plain_text, summary, source_notes, and download_count (downloads in the last 30 days, not lifetime)
• Fails with not_found or catalog_unavailable

gutenberg_get_text _tool

• Reads a book in offset/limit chunks (default limit 20,000, capped at 50,000), with the Gutenberg license header and footer stripped. requestedLimit and appliedLimit report both
• Prefers UTF-8 plain text and falls back to HTML converted to text; sourceFormat says which. Returns totalChars, length, remainingChars, hasMore, and provenance (ID, title, license URL). Use length, not limit, to compute the next offset
• find jumps to a passage instead of paging to it: the chunk starts at the first match at or after offset, and match reports its offset, length, and laterMatches. Matching ignores letter case and whitespace runs, and treats straight and curly quotes alike; accents and punctuation must match
• Converted HTML editions blank every line that holds no visible character and keep each table row on one line, with tab-separated cells
• Fails with not_found, audio_book (media_type "Sound"), no_text_format, offset_out_of_range, passage_not_found, text_file_not_found, text_fetch_rejected, invalid_text_response, text_fetch_failed, or catalog_unavailable

gutenberg_browse_popular _tool

• Returns up to 32 titles (limit, default 20), most downloaded in the last 30 days first. requestedLimit and appliedLimit report the cap
• Optional languages, topic, mime_type, and copyright filters work as in search. totalInCatalog gives the full count
• Output reports truncation and carries a continuation pointing at gutenberg_search_books. Empty results succeed with a notice; fails with catalog_unavailable

Features

Built on @cyanheads/mcp-ts-core: stdio and Streamable HTTP transports, pluggable auth (none / jwt / oauth), swappable storage (in-memory, filesystem, Supabase, Cloudflare KV/R2/D1), structured logging with optional OpenTelemetry tracing.

Gutenberg-specific:

• Searches and lookups read a local SQLite+FTS5 catalog mirror, falling back to Gutendex when the mirror is cold or missing a record
• Full plain-text retrieval from a Project Gutenberg content mirror, with UTF-8/HTML fallback
• Book text is cached per tenant, so later chunk reads skip the fetch
• No API key required

Agent-friendly output:

• is_readable on every search and browse result flags editions gutenberg_get_text can read
• offset, length, totalChars, remainingChars, and hasMore support reliable sequential reads
• provenance and sourceFormat on every text response cover attribution and fidelity

Getting started

Connect to the hosted instance, no install:

claude mcp add --transport http gutenberg-mcp-server https://gutenberg.caseyjhand.com/mcp

Or run it locally:

claude mcp add gutenberg-mcp-server -- npx -y @cyanheads/gutenberg-mcp-server

For other clients, add this to the MCP config file:

{
  "mcpServers": {
    "gutenberg-mcp-server": {
      "command": "npx",
      "args": ["-y", "@cyanheads/gutenberg-mcp-server"]
    }
  }
}

No API key required. For Docker or a source build, see Running the server.

Configuration

Build the local catalog with bun run mirror:init and check it with bun run mirror:verify. The server never harvests on startup, and uses Gutendex until the first harvest completes. Keep the catalog up to date with bun run mirror:refresh on a host scheduler, or set GUTENBERG_MIRROR_REFRESH_CRON.

Running the server

Local development

Requires Bun v1.4.0 or higher (or Node.js v24+).

• Clone and install:

  git clone https://github.com/cyanheads/gutenberg-mcp-server.git
  cd gutenberg-mcp-server && bun install
  cp .env.example .env   # optional overrides
  

• Build and run:

  # One-time build
  bun run rebuild
  
  # Run the built server
  bun run start:stdio
  # or
  bun run start:http
  

• Run checks and tests:

  bun run devcheck   # Lint, format, typecheck, security
  bun run test       # Vitest test suite
  bun run lint:mcp   # Validate MCP definitions against spec
  

Docker

docker run --rm -p 3010:3010 ghcr.io/cyanheads/gutenberg-mcp-server:latest

The image defaults to HTTP transport, stateless session mode, and logs to /var/log/gutenberg-mcp-server. To run it over stdio from an MCP client, use "command": "docker" with "args": ["run", "-i", "--rm", "-e", "MCP_TRANSPORT_TYPE=stdio", "ghcr.io/cyanheads/gutenberg-mcp-server:latest"]. To build locally, docker build -t gutenberg-mcp-server .; OpenTelemetry peer dependencies are installed by default — build with --build-arg OTEL_ENABLED=false to omit them.

Project structure

Development guide

See CLAUDE.md / AGENTS.md for development guidelines and architectural rules. The short version:

• Handlers throw, framework catches — no try/catch in tool logic
• Use ctx.log for request-scoped logging, ctx.state for tenant-scoped storage
• Register new tools via the entry arrays in src/index.ts
• Wrap external API calls: validate raw → normalize to domain type → return output schema; never fabricate missing fields

Contributing

Issues are welcome. Run checks and tests before submitting:

bun run devcheck
bun run test

Data attribution

Data from Project Gutenberg is in the public domain. Catalog metadata sourced from Gutendex (MIT license).

License

Apache-2.0 — see LICENSE for details.