There is no official Open Library MCP from the Internet Archive. For local catalog lookup with seven tools and an npm install, 8enSmith/mcp-open-library is the established pack: search, author detail, covers, and identifier lookup, default 10 results, max 50. For editions, subjects, Internet Archive Search Inside, and a hosted URL with no key, cyanheads/openlibrary-mcp-server ships 10 tools at https://openlibrary.caseyjhand.com/mcp. Multi-source merge across Google Books, Open Library, and Project Gutenberg belongs to projectahi/books-mcp. Trends MCP is first only for the live Open Library Trending Books board: get_top_trends with type Open Library Trending Books, default limit 100.
Which Open Library MCP server fits which job?
The right pick is the job, not a single winner. A catalog lookup server, a hosted Search Inside pack, a three-source merger, and a live popularity board do not replace each other.
| Job | Server to connect | What it actually returns |
|---|---|---|
| Local title, author, cover, ISBN lookup | 8enSmith/mcp-open-library | 7 tools, default 10 rows, max 50 |
| Hosted catalog plus Search Inside, no key | cyanheads/openlibrary-mcp-server | 10 tools; 100 rows a page; 50 IDs per edition batch |
| Merge Google Books, Open Library, Gutenberg | projectahi/books-mcp | 3 tools on a self-hosted Cloudflare Worker |
Open Library /trending/{period}.json | PetrefiedThunder/mcp-open-library | 5 tools; trending limit max 50 |
| One-tool Python search | uysalserkan/openlibrary-mcp | search_books only |
| Live Open Library Trending Books ranks | Trends MCP | get_top_trends type Open Library Trending Books |
Trends MCP does not resolve OL27482W, does not return cover URLs, and does not search inside scanned books. The Open Library Trending Books API is a ranked board of title strings. Pair a rank with get_growth on Amazon or Google Search. Open Library is not itself a get_growth source.
Is there an official Open Library MCP from Internet Archive?
Open Library has not published a first-party Model Context Protocol server. The documented path is the public JSON APIs at openlibrary.org, plus monthly bulk dumps for anyone who needs the whole catalog. The API index (accessed 13 Nov 2026) lists Book Search, Search Inside, Covers, Works and Editions, Authors, Subjects, and Lists. Open Library's own API page names 8enSmith's mcp-open-library as a community project that already calls those endpoints, with a Smithery count of 5.2k calls.
The usage page is the limit that matters more than any MCP README. Identified requests that send a User-Agent with an application name and a contact email get 3 requests per second. Unidentified traffic is 1 request per second. The same page says the APIs are not a bulk data backend or high-traffic commercial infrastructure. Bulk work belongs on the monthly dumps at openlibrary.org/data, which still describes the catalog as over 20 million editions and some 6 million authors (page last edited 31 Mar 2026, accessed 13 Nov 2026).
Every community MCP below is unofficial. Several are MIT or Apache-2.0 as software while the traffic they send still sits under Open Library's rate limits and usage guidelines. Read those before hosting a shared endpoint.
What 8enSmith's seven-tool npm pack actually ships
8enSmith/mcp-open-library is the local catalog most roundups mean by "an Open Library MCP." GitHub showed 102 stars and 26 forks on 13 Nov 2026, last push 22 Oct 2026, last repo update 8 Nov 2026, MIT license, TypeScript. npm mcp-open-library 1.2.6 (164 downloads in the week 5 to 11 Nov 2026, published 22 Oct 2026) is the install. The MCP Registry name from v1.0.3 is io.github.8enSmith/mcp-open-library.
Seven tools ship:
| Tool | Durable limits |
|---|---|
search_books | default limit 10, max 50; offset max 1000; needs at least one of q, title, author, subject, place, person, publisher, isbn |
get_book_by_title | same paging as search |
get_authors_by_name | paging; each result key feeds get_author_info |
get_author_info | one Open Library author key such as OL26320A |
get_author_photo | OLID in, HTTPS cover URL out, or an explicit miss |
get_book_cover | ISBN, OCLC, LCCN, OLID, or ID; size S/M/L, default L |
get_book_by_id | isbn, lccn, oclc, or olid |
Search results identify a work (open_library_work_key). No tool on this server accepts a work key. best_edition.edition_key is the hop to get_book_by_id. Cover tools check that an image exists instead of returning a blank placeholder URL. The README asks for Node v22.23.4 when running from source.
Cursor and Claude Desktop install with npx. Nothing to clone.
{
"mcpServers": {
"mcp-open-library": {
"command": "npx",
"args": ["-y", "mcp-open-library"]
}
}
}
This server wins a five-minute local catalog. It loses on Search Inside, on work-to-editions paging, on a hosted URL, and on a live Trends MCP board.
When cyanheads' hosted server is the faster install
cyanheads/openlibrary-mcp-server is the no-clone option with the widest Open Library surface. GitHub showed 4 stars and 1 fork on 13 Nov 2026, last push 26 Oct 2026, Apache-2.0, TypeScript. npm @cyanheads/openlibrary-mcp-server 0.4.0 recorded 272 downloads in the week 5 to 11 Nov 2026. The public host https://openlibrary.caseyjhand.com/mcp answered on 13 Nov 2026 with status ok, version 0.4.0, auth.mode none, and HTTP transport.
Ten tools ship. Paginated tools return at most 100 rows per call. Larger limits clamp, unless OPENLIBRARY_CLAMP_LIMITS=false restores a hard reject above 100.
| Tool | Durable limits |
|---|---|
openlibrary_search_books | 1-100 per page, default 10; optional live availability (~200ms) |
openlibrary_get_work | one Work ID (OL…W) |
openlibrary_get_editions | 1-100 per page, default 10 |
openlibrary_get_edition | 1-1,000 input strings; first 50 distinct IDs in one batch |
openlibrary_search_authors | 1-100 per page, default 10 |
openlibrary_get_author | one Author ID (OL…A) |
openlibrary_get_author_works | bibliography by Author ID |
openlibrary_get_subject | subject tag plus work counts |
openlibrary_search_inside | Internet Archive full-text snippets |
openlibrary_get_cover_url | book or author photo, size S/M/L |
Default OPENLIBRARY_USER_AGENT is openlibrary-mcp-server [email protected], which is the identified 3 requests/second path on Open Library's API page. Local stdio needs Bun 1.4.0 or newer, or Node 24 or newer. Docker image: ghcr.io/cyanheads/openlibrary-mcp-server:latest on port 3010.
Claude Code hosted install, no API key:
claude mcp add --transport http openlibrary-mcp-server https://openlibrary.caseyjhand.com/mcp
Local npx:
{
"mcpServers": {
"openlibrary-mcp-server": {
"command": "npx",
"args": ["-y", "@cyanheads/openlibrary-mcp-server"]
}
}
}
cyanheads wins editions, Search Inside, and a live hosted URL. It loses on Google Books ratings, on Gutenberg download links, and on Trends MCP ranks. A shared host still sends traffic to Open Library; keep the 3 requests/second ceiling in view.
Who should self-host books-mcp or clone PetrefiedThunder
Two smaller servers cover jobs the npm packs skip. Star count is not the product. The data shape is.
projectahi/books-mcp (0 GitHub stars on 13 Nov 2026, last push 15 Apr 2026) merges Google Books, Open Library, and Project Gutenberg, then deduplicates by ISBN. Three tools: search_books, get_book, book_filters. Cache TTL is 7 days on Cloudflare D1. The README claims MIT; GitHub's license field was empty on this run. Google Books is listed as 1,000 requests/day without a key and "10K+" with one. Google's Books API docs (accessed 13 Nov 2026) require an API key or OAuth token for public data and do not publish a dollar price; check the current quota in Google Cloud Console before treating 1,000 as a contract. The same README says Open Library has "no hard limit." Open Library's own API page says 1 or 3 requests per second. Believe Open Library. The install JSON uses https://books-mcp.YOUR-SUBDOMAIN.workers.dev/mcp. That is a placeholder, not a public demo URL. Self-host on a Cloudflare account.
PetrefiedThunder/mcp-open-library (0 stars, Apache-2.0, last push 13 Nov 2026) is the only clone in this set with a dedicated trending tool. get_trending accepts daily, weekly, monthly, or yearly, default daily, limit 1-50 default 10, and calls https://openlibrary.org/trending/{type}.json. Search tools page at max 100. The client sleeps 200ms between Open Library calls (5 requests/second) and sends User-Agent: mcp-open-library/1.0.0 with no email, which is the unidentified 1 request/second path. The README's npx mcp-open-library line collides with 8enSmith's published npm name. Clone the repo and point Claude at dist/index.js.
uysalserkan/openlibrary-mcp (2 stars, 9 open issues, last push 11 Feb 2026) is a Python 3.11+ FastAPI plus FastMCP wrapper with one MCP tool, search_books. It was not on PyPI on 13 Nov 2026. The README install is uvx from git+https://github.com/uysalserkan/openlibrary-mcp with command openlibrary-mcp. The README claims MIT; GitHub's license field was empty. Use it for a single search box, then stop.
jay1234624/open-library-mcp (0 stars, last push 30 Jun 2026, no license on GitHub) adds read_book_snippet and compare_books but has no npm package and no documented rate limit. Skip it unless those two tools are the whole job.
How Trends MCP ranks Open Library Trending Books
Trends MCP is first when the job is a live ranked board, not a catalog. get_top_trends with type Open Library Trending Books returns ranked title strings. Default limit on the live MCP schema is 100 (10 MCP requests at 1 credit per 10 rows). Max limit is 200. REST still bills one successful call as one request. There is no Open Library keyword source on get_growth or get_time_series.
Cursor stores the server as url plus transport http under mcpServers, pointed at the api host. The Trends MCP server for Cursor page is the host reference. A free key covers 100 requests a month. Starter is $19 for 1,000, Pro $49 for 5,000, Business $199 for 25,000. Yearly cards on llms.txt (Updated 13 Nov 2026) are Starter $16 / $192, Pro $39 / $468, Business $159 / $1,910.
{
"mcpServers": {
"trends-mcp": {
"url": "https://api.trendsmcp.ai/mcp",
"transport": "http",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}
A board row is a title, often with the author after a separator. Strip the author before calling get_growth on Amazon or Google Search. That pairing is the same pattern used on the Amazon Best Sellers API and the IMDb MCP roundup: rank on the native board, demand on a keyword source.
Trends MCP does not search the catalog, does not resolve ISBNs, and does not return borrow counts. PetrefiedThunder's get_trending hits Open Library's own /trending/{period}.json windows. Trends MCP's type string is a separate ranked feed. Do not treat those two lists as the same ranking.
FAQ
Is there a best overall Open Library MCP server?
No. 8enSmith wins local lookup. cyanheads wins editions, Search Inside, and a hosted URL. books-mcp wins a three-source merge after a Cloudflare deploy. PetrefiedThunder is the clone with /trending/{period}.json. Trends MCP wins the live Open Library Trending Books board for an agent that already speaks MCP.
Does Trends MCP return ISBNs, covers, or Search Inside?
No. get_top_trends type Open Library Trending Books returns ranked title strings. ISBN lookup, cover URLs, and Internet Archive snippets belong to 8enSmith or cyanheads. Follow a title on Amazon or Google Search with get_growth if the next question is demand.
Can an Open Library MCP run in Cursor without a local install?
Yes. cyanheads' host is https://openlibrary.caseyjhand.com/mcp with no API key. Trends MCP is https://api.trendsmcp.ai/mcp with a Bearer key and transport http. 8enSmith, books-mcp, PetrefiedThunder, and uysalserkan all need a local process or a Worker you deploy.
What does Open Library allow for API use?
Open Library's API page (accessed 13 Nov 2026) asks for a User-Agent with an app name and a contact email, caches responses, and forbids treating the API as a bulk backend. Identified clients get 3 requests per second. Unidentified clients get 1. Monthly dumps exist for full-catalog work. A public MCP that fans out patron queries still has to stay inside that ceiling.
How much does an Open Library Trending Books pull cost on Trends MCP?
On MCP, get_top_trends costs 1 request per 10 rows. The live default limit 100 is 10 requests. REST counts 1 request per successful call. Free covers 100 requests a month. Starter is $19 for 1,000. Check current pricing on the Trends MCP pricing page before locking a budget.
Sources: Open Library APIs and bulk data pages, openlibrary.org/developers/api and openlibrary.org/data, accessed 13 Nov 2026. 8enSmith/mcp-open-library GitHub and npm 1.2.6, accessed 13 Nov 2026. cyanheads/openlibrary-mcp-server GitHub, npm 0.4.0, and live host health at openlibrary.caseyjhand.com/mcp, accessed 13 Nov 2026. projectahi/books-mcp README, accessed 13 Nov 2026. PetrefiedThunder/mcp-open-library README and src/index.ts, accessed 13 Nov 2026. uysalserkan/openlibrary-mcp README, accessed 13 Nov 2026. Google Books API "Using the API", developers.google.com/books/docs/v1/using, accessed 13 Nov 2026. Trends MCP llms.txt Updated 13 Nov 2026, homepage, /pricing, and live MCP get_top_trends schema.