macOS  ·  App Store + Direct Download

Private AI for
your documents.

Search, chat, and explore your files with a local AI workspace that never sends a byte to the cloud. Hybrid retrieval across 91+ formats, native OCR, and cited answers — all on your Mac.

Try Free for 14 Days See How It Works

macOS 12+  ·  Apple Silicon & Intel  ·  No account required

PaperMind app showing document search and AI chat with cited answers
91+
File formats indexed
4 OCR
Engines including Apple Vision
Hybrid
BM25 + Vector + Reranking
Zero
Data sent to the cloud
How It Works

Import. Search. Understand.

Three steps to turn any pile of documents into a searchable, conversational knowledge base — entirely on your Mac.

1. Import

Drop a folder or pick files. PaperMind parses PDFs, Word docs, spreadsheets, presentations, code, images — 91+ formats. OCR kicks in automatically for scanned pages.

2. Search

Hybrid retrieval combines keyword precision with semantic understanding. Type a question or a phrase — PaperMind finds the right passages even when wording doesn't match.

3. Ask AI

Have a conversation grounded in your documents. PaperMind cites exact sources, so you always know where an answer came from. Local models or cloud — your choice.

Who Uses It

Built for anyone who works with documents

No special setup. No technical knowledge required. Drop your files in and start asking questions.

📚
Researcher
Search across hundreds of papers

Index your entire literature library. Find specific claims, compare findings across papers, and get cited answers without digging through PDFs one by one.

⚖️
Lawyer / Paralegal
Search case files with citations

Upload contracts, briefs, and precedents. Ask plain-language questions and get answers with exact page references — the kind you can actually cite.

💼
Consultant
One workspace per client

Separate workspaces keep client projects isolated. Search through proposals, reports, and deliverables without mixing up context between engagements.

📰
Journalist
Search transcripts and records

Scanned FOIA documents, interview transcripts, background research — OCR makes anything legible searchable. All offline, nothing leaving your machine.

Features

Everything you need, nothing you don't

Document Import

If you can read it, PaperMind can index it

91 formats supported natively — no plugins, no third-party converters. Drop a folder and everything inside gets parsed automatically. Scanned pages and photos of documents trigger OCR without any manual step.

  • PDF, DOCX, XLSX, PPTX, EPUB, HTML, Markdown, CSV, and 83 more
  • 4 OCR engines including Apple's native Vision framework
  • Automatic scanned page detection — no manual flagging
  • Folder watching — add files and they get indexed automatically
Drop folder or click to import PDF DOCX XLSX PPTX +87 more Indexing 3 of 4 files… 75%

Hybrid Search

Finds what you mean, not just what you typed

Most document search fails when your words don't match the document's words. PaperMind runs keyword and semantic search in parallel, then merges results using Reciprocal Rank Fusion — better results than either approach alone.

  • BM25 index catches exact terminology and proper nouns
  • Dense vector search catches meaning and synonyms
  • Query expansion adds related terms to cast a wider net
  • Cross-encoder reranker puts the most relevant passages first
What were the key findings on hybrid retrieval? RESEARCH PAPER · P. 12 Hybrid retrieval outperforms dense-only search by 23% on recall@10 and 18% on precision for domain-specific queries… LITERATURE REVIEW · P. 34 BM25 excels at exact term matching while vector search captures semantic similarity. RRF fusion consistently improves both metrics… CONFERENCE NOTES · P. 7 Cross-encoder reranking on top of hybrid search adds another 11% precision improvement on ambiguous queries…

Conversational AI

Every answer comes with a source

Ask questions in plain language. PaperMind runs the full retrieval pipeline, passes the most relevant passages to the language model, and requires citations in every response. You always know which document and page an answer came from.

  • Page-level citations in every AI response
  • Works with Ollama, llama.cpp, OpenAI, and Anthropic Claude
  • Switch models anytime — no lock-in
  • Local models work fully offline, no internet needed
You
What are the performance gains from adding a reranker?
PaperMind
Adding a cross-encoder reranker pushes precision up an additional 11%, with the largest gains on ambiguous and multi-hop queries.
Research Paper, p.18
You
Does the improvement hold on shorter documents?
Format Support

91+ formats. No plugins.

Every major document type parsed natively. Scanned documents handled automatically through four OCR engines.

PDF DOCX XLSX PPTX Markdown HTML CSV EPUB YAML TOML JSON Source Code Images (OCR) RTF Plain Text + 76 more

Scanned documents, photographed receipts, old contracts — PaperMind ships with four OCR engines including Apple's native Vision framework. If the text is legible, it gets indexed.

Under the Hood

Enterprise-grade retrieval on your laptop

PaperMind runs a four-stage retrieval pipeline — the same techniques used in professional search infrastructure, running entirely on-device.

1
Hybrid Search
Your query hits both a BM25 keyword index and a dense vector store simultaneously. Results from both are merged using Reciprocal Rank Fusion — giving you the precision of keyword search and the recall of semantic search.
BM25 + Dense Vectors + RRF
2
HyDE + Query Expansion
PaperMind generates a hypothetical answer to your question and uses it to improve the vector search pass. Your query is also expanded with related terms to catch relevant passages that don't use your exact wording.
Hypothetical Document Embeddings
3
Cross-Encoder Reranking
A dedicated reranking model scores every candidate passage against your full query. The most relevant passages rise to the top — dramatically improving precision on ambiguous or multi-part questions.
+11% precision over hybrid alone
4
Grounded Answer with Citations
The language model receives only the highest-scoring passages and generates a response with page-level citations. Every claim is traceable to a specific document and page number.
Fully cited, always verifiable
Privacy First

Your documents never leave your Mac

This isn't a marketing claim — it's how the architecture works. There are no servers to send your files to.

Offline by default

Parsing, OCR, embedding, indexing, and search all run on your machine. No internet needed. No cloud accounts. No telemetry pinging home.

Zero data collection

Apple's PrivacyInfo manifest confirms it: no tracking, no analytics, zero declared data types collected. We don't know what you index and we don't want to.

Offline licensing

License verification uses local cryptographic key signing — no activation server, no phone-home. Works on air-gapped machines and behind corporate firewalls.

Why PaperMind

Not all document AI is equal

The privacy question is the obvious one. But there's more to it.

PaperMind ChatGPT upload Notion AI
Files stay on your device ✓ ✕ ✕
Works fully offline ✓ ✕ ✕
Page-level citations ✓ Sometimes ✕
91+ file formats ✓ PDF, DOCX only Limited
Hybrid retrieval + reranking ✓ Unknown ✕
Scanned document OCR ✓ Basic ✕
Pricing model $49 once $20/month $16+/month

Comparison reflects publicly available information as of April 2026.

Workspaces

One app, as many knowledge bases as you need

Create separate workspaces for different projects, clients, or research areas. Each has its own index, chat history, and settings.

📚
Research Library

Papers, literature reviews, and notes in one searchable collection.

⚖️
Case Files

Contracts, briefs, and precedents with exact page citations.

💼
Client Projects

Proposals, reports, and correspondence — switch context in a click.

🛠️
Technical Docs

API documentation, specs, and source code — all indexed and searchable.

Available in 10+ languages: English, French, German, Spanish, Portuguese, Japanese, Korean, Italian, Dutch, Arabic, Chinese.

Pricing

Pay once, own it forever

No subscriptions. No recurring charges. A 14-day free trial with full access — no card required.

Free Trial

$0

14 days, full access, no card required.

  • All features unlocked
  • All file formats
  • Hybrid retrieval + reranking
  • All OCR engines
Start Trial

Personal

$49 once

One-time purchase. Lifetime updates.

  • 3 workspaces
  • 500 documents per workspace
  • Dense retrieval
  • 2 LLM providers
  • Email support
Buy Personal
Best Value

Professional

$99 once

One-time purchase. Everything unlocked.

  • Unlimited workspaces
  • Unlimited documents
  • Hybrid retrieval + HyDE + reranking
  • All 4 OCR engines
  • All LLM providers
  • All parsers & formats
  • Priority support
Buy Professional

Launch discount: 30% off for the first 500 customers.  ·  Also available on the Mac App Store.

FAQ

Common questions

No. Every document you import is processed, indexed, and stored entirely on your Mac. PaperMind works offline by default. The only time any data leaves your machine is if you explicitly enable a cloud LLM — and even then, only the relevant text passages are sent, never full documents.
91+ formats including PDF, Word (.docx), Excel (.xlsx), PowerPoint (.pptx), Markdown, HTML, EPUB, CSV, YAML, TOML, source code files, and images (via OCR). Scanned documents are handled automatically by four built-in OCR engines.
Hybrid retrieval combines keyword search (BM25) with semantic vector search and merges the results using Reciprocal Rank Fusion. This means PaperMind finds answers whether you search by exact terms or by meaning — significantly more accurate than either method on its own. The Professional plan adds HyDE and cross-encoder reranking on top of that.
PaperMind supports Ollama for running local models (Llama, Mistral, Gemma, Phi, and any Ollama-supported model), llama.cpp for embedded GGUF models, and cloud providers including OpenAI and Anthropic Claude. You can switch between models at any time with no lock-in.
No. Document processing, OCR, indexing, and search all run locally. You only need internet if you choose a cloud LLM for the Ask AI feature. Ollama runs local models offline. PaperMind also works on air-gapped machines and behind corporate firewalls.
Personal gives you 3 workspaces, 500 documents each, dense-only retrieval, and 2 LLM providers — good for personal use. Professional unlocks unlimited workspaces and documents, hybrid retrieval with HyDE and cross-encoder reranking, all 4 OCR engines, all LLM providers, and priority support. Both are one-time purchases.
No. PaperMind is a one-time purchase. Personal is $49, Professional is $99. Pay once, own it forever. There's also a 14-day free trial with full feature access — no credit card required.
Not at all. Drop files in, search, ask questions — that's it. The retrieval pipeline runs automatically in the background. For technically-minded users who want to dig deeper, there's an Expert Mode with a performance dashboard and fine-grained pipeline controls.
After 14 days, PaperMind locks access until you purchase a license. Your data and workspaces are preserved — nothing is deleted. Once you buy a license, everything unlocks immediately and you're set for life.
PaperMind is macOS-only right now (Apple Silicon and Intel, macOS 12+). Windows and Linux builds are planned for a future release. Sign up for our newsletter to get notified when cross-platform support launches.

Ready to try PaperMind?

14-day free trial. No account, no credit card, no cloud.

Download on App Store Direct Download (.dmg)
macOS 12+ Apple Silicon & Intel 14-day trial No account