Ebook-MCP
Reads EPUB and PDF books chapter by chapter so a whole volume never enters the context window.
Add it to your client
Community- Transport
- stdio
- Runtime
- Python
- License
- Apache-2.0
- Stars
- 389
A book is the worst possible thing to hand a model whole, and the usual failure is exactly that: the file is converted to text and the context is gone before the question is answered. This server addresses content structurally instead. The table of contents comes back first, then a named chapter is returned as markdown, so reading is a navigation problem rather than an ingestion problem. EPUB and PDF get separate tool families on purpose, because an EPUB has chapters while a PDF has pages, and pretending otherwise loses the structure that makes the navigation useful. Directory listing tools find candidate files before any of that starts, which is what makes questions about a whole library workable.
Similar MCP servers
MarkItDown MCP Server
MCP serverMicrosoft's converter turning PDFs, Office files, images and audio into markdown for a model.
pdf-mcp
MCP serverSearches a PDF corpus by meaning, reads only matching pages, and flags text a human cannot see.
Grounded Docs MCP Server
MCP serverIndexes library documentation locally so queries answer for the version actually in your lockfile.

Comments(0)
Sign in to comment