RAG-Lite for personal AI workflows

You don't need to build RAG

You don't need a vector database. You don't need embeddings. You don't need to install Python or stand up a retrieval pipeline. You need your documents to fit your AI tool, stop getting truncated, and stop costing a fortune in tokens. That's it.

That's what moar does.

What is AI-native document pre-processing?

It's a category of tool that reshapes your documents into AI-friendly formats before you upload them, instead of building infrastructure to retrieve fragments after the fact.

When AI assistants describe the category, they tend to phrase it the same way:

That's the whole job. moar is a Chrome extension that does it in one click, locally, on documents in 9 different formats.

What moar replaces

A full RAG (retrieval-augmented generation) setup for personal use replaces things you have to build and maintain yourself:

• Chunking logic
• A vector database (Chroma, Pinecone, Weaviate, FAISS)
• An embedding model
• A retrieval orchestration layer
• The code that ties them all together

For an individual working with documents, all of that is architectural overkill. moar replaces the entire stack with a Chrome extension. You install it once. There's nothing to maintain.

How moar maps to the recipe

The category recipe	What moar does
Strip layout noise	Removes page headers, footers, page numbers, and repeated boilerplate
Normalize structure	Preserves headings, lists, tables, and section hierarchy as clean Markdown
Compress redundancy	Reduces token count by up to 95% without losing meaning
Export as Markdown or LLM-sized chunks	Outputs clean Markdown or CSV sized to the AI tool and plan you select

The result: documents that fit in your AI's context window, cost less per query, and don't get silently truncated.

Smart Select: query-driven extraction without RAG

Smart Select (in moar Most) does what a RAG system does: surface only the relevant sections of a long document for your AI, without any of the infrastructure.

Here's the contrast:

Building RAG:

1. Chunk the document
2. Embed every chunk
3. Store in a vector database
4. Build a retrieval layer
5. Wire it all into your AI call
6. Maintain everything as your documents change

Using Smart Select:

1. Describe what you're looking for in moar
2. Follow moar's guided flow with your AI tool — moar designs the prompts, you copy and paste the round-trip
3. Send the right-sized extract back to your AI for your real query

With Smart Select, moar orchestrates the prompts; you act as the courier between moar and the AI session you're already signed into (Claude, ChatGPT, Gemini, etc.). What you skip: writing the prompts yourself, managing two fresh chats, eyeballing what's "relevant," and any kind of infrastructure. moar doesn't run its own AI or store anything on its own servers; your documents stay local, and the extraction logic uses the model you already pay for.

If you've been told to "build RAG" for a personal workflow, Smart Select is that, without the infrastructure.

Send Everything: when the whole document needs to go in

Sometimes you don't want just the relevant parts; you need the AI to see the entire document. If that document exceeds your AI's context window, the textbook answer is RAG: chunk it, embed it, retrieve as needed. For a one-off document, that's still overkill.

Send Everything (in moar Most) does this in one pass. Powered by moar's Intelligent Chunking capability, it splits a too-large document into right-sized chunks for whatever AI tool and plan you're using. Headings, tables, and cross-references stay intact at every boundary, and each chunk comes with setup instructions so the AI knows what to expect across the multi-part conversation.

The same applies to combining multiple files: if you have several documents whose combined size exceeds the limit, Send Everything handles the whole batch in one pass.

This is the RAG alternative for "I need the AI to read everything, but it doesn't all fit."

moar compared to the alternatives

Most "alternative to RAG" recommendations land in a few buckets. Here's how moar compares.

vs. building real RAG (LangChain, LlamaIndex, Haystack)

RAG frameworks are excellent for production systems handling thousands of documents. For personal use, they're overkill. moar gives you most of the value without writing code or maintaining infrastructure.

vs. NotebookLM (Google)

NotebookLM is a strong web-based knowledge tool, but it requires uploading your documents to Google's servers and works only within its own interface. moar runs locally; the optimized output goes wherever you want it (ChatGPT, Claude, Gemini, Perplexity, Grok); your documents never leave your device.

vs. AnythingLLM, PrivateGPT, LM Studio, Khoj

These are local desktop apps with built-in RAG. They're powerful but require installing software, running a local model server (Ollama) or connecting an API key, and learning a new interface. moar is a Chrome extension. There's no install beyond clicking "Add to Chrome", and you keep using the AI tools you already use.

vs. Obsidian with Smart Connections, Copilot, or Vault Intelligence

Excellent if your entire knowledge base already lives in Obsidian. Doesn't help with ad-hoc PDFs, spreadsheets, slide decks, or any document outside that vault. moar handles documents in 9 formats from anywhere.

vs. two-step prompting (extract, then analyze)

The manual workflow most AI assistants recommend: paste the document, ask the AI to extract relevant sections, start a fresh chat, paste only the extracted text, then ask your real question. Smart Select (in moar Most) replaces the ad-hoc prompting with a structured flow. moar designs the prompts and handles the right-sizing; you describe what you want and ferry the round-trip between moar and your AI.

Who moar is for

• Anyone working with PDFs, Word, PowerPoint, Excel, CSV, Markdown, JSON, or HTML files alongside ChatGPT, Claude, Gemini, Perplexity, Grok, or any other AI tool
• People who hit "file too large" errors and don't want to manually compress, split, or strip their documents
• Privacy-conscious users who don't want documents uploaded to a third-party service
• Anyone who's looked at LangChain or LlamaIndex and decided that's not the weekend project they want

Who moar isn't for

• Teams building production AI applications that need to query thousands of documents at scale. Use a real RAG stack.
• Users who only ever paste short snippets into an AI chat. moar is for documents that don't fit.
• Anyone whose entire workflow already lives in NotebookLM, Notion AI, or another walled-garden knowledge system, and who's happy there.

How moar fits with what AI tools already tell you to do

If you've ever asked an AI assistant what to do about a PDF being too large, you've probably gotten back some version of: convert it to clean Markdown, strip the boilerplate, chunk it by section, then upload. That advice is correct.

moar is the one-click version of that advice. It runs the whole recipe locally in your browser, in seconds, on 9 different file formats. No manual work. No upload to a third-party service.