tale.fyi

tale for AI assistants

Find any passage in a book and link straight to it.

Tale gives AI assistants exact, shareable links to paragraphs in its books, with the paragraphs around them for context. The links open the same editions people read on Tale. No account is needed.

Any assistant that can open a URL can ask with a plain web request. An MCP connection can do the same, and can also read books, browse shelves and make reading plans.

ask with a web request

Find a book, or a tale a reader wrote, by its title or author:

https://tale.fyi/api/tales?q=dracula

Then search it for words it contains:

https://tale.fyi/api/passages?path=dracula&q=Demeter

The answer is JSON. Each match has the paragraph’s text, its chapter, the paragraph before and after it, and a link such as tale.fyi/dracula#chapter-7-p16. Matches come in reading order. path is the part of the address after tale.fyi/, and q is the tale’s own words. The full description, with the response’s fields, is in llms.txt.

connect in ChatGPT

  1. In ChatGPT on the web, enable Developer mode in Settings → Security and login. Your plan or workspace needs to allow custom MCP connections.
  2. Open ChatGPT Plugins, choose the plus button, and add a connection named Tale. Use this URL: https://tale.fyi/api/mcp
  3. Review the tools ChatGPT discovers. In a new chat, select Tale from the tools menu and ask for an author's tales, a passage, or a reading plan.

ChatGPT's menus and availability can change. Its current connection guide has the latest steps. Tale itself requires no account for these public tools.

browse a shelf

Give your assistant a Tale author link, such as tale.fyi/@eliza, and ask which tales are on that page. It can also find and list books by author, genre, era, or curated list. Large shelves arrive in pages with links to each tale.

try a passage

Ask your assistant:

Find the paragraph in Dracula containing “To My Dear Friend Hommy-Beg” and give me its Tale link.

Tale's search returns the verified paragraph marker dedication-p1 and this passage link. Search matches text within one chosen tale; it is not a search across every book.

send a request yourself

Other MCP clients can use Tale's Streamable HTTP endpoint at https://tale.fyi/api/mcp. A client handles initialization and tool discovery. This direct, read-only request works on its own:

curl -sS https://tale.fyi/api/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  --data '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"search_passages","arguments":{"path":"dracula","query":"To My Dear Friend Hommy-Beg"}}}'

The response's result.structuredContent.matches list contains the paragraph ID, text, and a Tale URL. For the request above, the first match is tale.fyi/dracula#dedication-p1.

reading and making plans

Ask for a chapter or an inclusive passage range, or read a whole tale. Long books arrive in bounded pages, each with a continuation cursor. The assistant can use exact paragraph markers from those pages to build a dated, annotated reading plan.

Publishing a plan creates a permanent public link. Review its dates and notes before asking the assistant to publish it. Reading progress stays with the reader, as it does elsewhere on Tale. Connecting Tale does not change how ordinary web search fetches its pages.

This connection cannot create, edit, or publish tales. To write a tale, use Tale's editor.