Token Saver: open-source MCP extension cuts Claude's token use on PDFs by 90-99%

On July 30, 2026, Marktechpost released Token Saver, a free open-source extension for MCP (Model Context Protocol) in Claude Desktop. It cuts token use when analyzing PDFs by 92-99% through local Hybrid RAG, and it does not upload the file itself to the model's servers.

Why large PDFs eat tokens

Large PDFs are expensive in chat because the entire conversation history is resent to the model on every turn, so the document is paid for again with each new question. According to Marktechpost, Claude by default renders every page as an image to preserve charts and layout, and separately extracts the text; the text alone takes 1500 to 3000 tokens per page.

Prompt Caching and Claude Projects soften the blow but do not solve the root problem: the whole document still crosses the boundary and lands on the provider's servers. If you need only two paragraphs from a 1000-page textbook, making an LLM comb through all 1000 pages is inefficient and raises the risk of hallucinations.

How Local Hybrid RAG works

Token Saver runs as a local MCP server, a lightweight background program on your machine that indexes the PDF and hands the model only the relevant excerpts. Instead of pushing all thousand pages through the model, hybrid search finds what's needed locally. That is cheaper, and it lowers the chance of fabricated answers.

• Release date: July 30, 2026, version v1.0, MIT license
• Token reduction: 92-99%
• Developer: Arnav Rai, a CS student at Rochester Institute of Technology, built during an internship at Marktechpost AI Media Inc
• Project supervisors: Jean-marc Mommessin and Asif Razzaq
• Setup requires neither Python environments nor terminal work

The file never leaves the device, so privacy is enforced by the architecture rather than by promises. The extension is distributed under the MIT license and is already available in version v1.0.

"Token use is cut by 92-99%, privacy is guaranteed, and setup requires exactly zero

Python environments and terminal configurations," according to the Marktechpost AI Media announcement.

What this means

Local RAG is becoming a practical way to work with large documents through cloud LLMs without overpaying for tokens on every follow-up question and without sending source files off the machine. For developers and researchers who regularly work through bulky PDFs, this is a direct saving, both on the model bill and on data privacy.

Frequently asked questions

What is Token Saver?

Token Saver is an open-source MCP extension for Claude Desktop under the MIT license (version v1.0). It processes PDFs locally and passes only the relevant excerpts to the model, cutting token use by 92-99%.

Do I need Python to install it?

No. According to Marktechpost, setup requires neither Python environments nor terminal configuration; the extension runs as a background program on your machine.

How much does Token Saver cost?

The tool is free and open: it is distributed under the MIT license, and the PDFs themselves are processed on the device without being uploaded to the provider's servers.