Find connections
across your files

FileGrind breaks documents into pieces and shows you how they connect. Which pages in different books talk about the same concept? Which research papers cluster around the same topic? One search, all your files, regardless of format.

Everything runs on your Mac. Local AI models, local search index, no cloud uploads. Your files stay yours.

We'll email you when it's ready. No spam.

What FileGrind does

Most apps handle one file at a time. Your PDF reader shows you one PDF. Your ebook app shows you one EPUB. They don't know about each other.

FileGrind grinds files down into structured pieces called chips. A chip might be a page from a PDF, a chapter from an EPUB, or a section from a text file. Once everything is chips, you can search across all of them, find which ones are related, and see connections that weren't visible before.

Chips: structured pieces

When you add a file to FileGrind, it gets broken down into chips. Each chip carries the content plus metadata: where it came from, what page or section, what type of content it is.

Chips are the unit of search and connection. You don't search files—you search chips. You don't link files—you link chips. This lets you be precise: link page 45 of one book to chapter 3 of another.

Example: Grinding a PDF
Input
machine-learning-textbook.pdf (400 pages)
Output
400 chips, one per page, each with extracted text and images
Plus
Vector embeddings for semantic search, auto-generated keywords

Local AI, no API keys

FileGrind runs AI models on your Mac using MLX. Embedding models for semantic search. Classification models for organizing. Summarization if you want it.

No API keys. No usage limits. No sending your documents to a server. Download a model once, run it forever. Works offline.

Available Models
Embeddings
all-MiniLM-L6-v2 for semantic search
LLM
Mistral-7B for analysis and summarization
More
Browse and download from HuggingFace
Built-in Capabilities
PDF text extraction EPUB parsing Image thumbnails Metadata extraction
Community Capabilities
Audio transcription coming
3D model metadata coming
Code parsing coming

Capabilities: what you can do to files

A capability is an operation you can perform on your files. Extract text from a PDF. Parse an EPUB's structure. Generate embeddings for search.

FileGrind ships with capabilities for common file types. More get added through plugins. The capability system uses CAPNS for discovery and matching—your system finds the right tool for each file type automatically.

Plugins for new file types

We're launching with PDFs, EPUBs, text files, and images. But FileGrind isn't limited to these.

Plugins are standalone binaries that add capabilities. Want to grind 3D models? Audio files? Spreadsheets? Install a plugin or write your own. Plugins run in a sandboxed process and communicate via a simple protocol.

Plugin Architecture
Bundled
pdfczar, txtczar (text and markdown)
Installable
Download signed .pkg installers
Protocol
JSON in/out via stdin/stdout

Not a file manager, not a reader

FileGrind doesn't replace your PDF reader or ebook app. It doesn't manage where your files are stored.

It's a layer that sits between your files and shows you what's in them. Double-click a result and it opens in Preview, Acrobat, or whatever app you've set for that file type. FileGrind finds the connections. Your other apps do what they do best.

Your files stay on your Mac

No cloud uploads. No "processing on our servers." AI models run locally via MLX. Search indexes are stored locally. Unplug your internet—FileGrind keeps working.

✓ Local AI (MLX)
✓ Local storage
✓ Works offline

Beta: what's ready, what's not

FileGrind is launching in beta. Here's what works now and what's coming.

Working: PDF grinding, EPUB parsing, text files, image thumbnails, semantic search, local LLM analysis, the plugin system.

Coming: More file types via community plugins, better UI for managing blocks of chips, export options, and performance improvements for large libraries.