Local FAISS MCP server

This is a small helper that lets your AI assistant read your own files and answer questions about them. You feed it documents like PDFs, text files, or Word files, and later you ask questions in plain language. It is handy for anyone who has a pile of reports, manuals, or notes and wants quick answers without uploading anything to the internet.

What this MCP server does

You point this helper at your documents, and it reads them, splits them into small pieces, and stores them in a searchable index on your computer. When you ask your AI a question, the AI uses this helper to find the most relevant pieces from your files. The helper hands those pieces back, and your AI writes an answer based on them. If you turn on re-ranking, it does a second pass to put the best matches first. Everything is saved to disk, so you only need to add each document once.

Click to zoom

What you can do with it

• Add PDFs, text files, and Markdown files to a searchable library
• Ask questions in plain language and get answers based on your own documents
• Search across many files at once instead of opening them one by one
• Add Word, HTML, or EPUB files if you also install pandoc
• Get a short summary of what your documents say about a topic
• Get answers with the source file named, so you can check where it came from
• Index a whole folder of documents from the command line

Try asking your AI

• “Add the file ./reports/annual_review.pdf to my document store”
• “Search my documents for how our refund policy works”
• “What do my notes say about the project timeline?”
• “Summarize what my documents say about onboarding new staff”

What it gives back to you

You get back short pieces of text from your own documents, along with the file they came from and a relevance score. Your AI then turns those pieces into a written answer in the chat, often with the source file mentioned. If you use the built-in prompts, it can also give you a clean answer with citations or a short summary. Nothing is changed in your original files.

Before you start

Who it's for

Anyone who works with a lot of documents, like researchers, analysts, writers, or support staff, and wants quick answers from their own files without sending them anywhere.