b

Your clipboard,
but smarter.

BuffBrain remembers what you copied — and understands what it means. On-device semantic search finds the clip you're looking for, even when you don't remember the exact words.

Download View on GitHub
🧠 On-device AI 🔒 Zero telemetry ⚡ Native Rust 🆓 Free & MIT

Everything a clipboard manager should be.

And a few things no clipboard manager has ever been.

Semantic Search

Search by meaning, not just text. A MiniLM-L6 model runs on-device to embed every clip into a 384-dimensional vector. Type a concept, get matching results.

🔌

Zero Cloud

Everything runs locally. The ONNX model is bundled (~22MB). No data ever leaves your machine. No account. No API key required.

🧪

Auto-Classification

30+ heuristic rules detect code, links, and text instantly. Optionally connect OpenRouter for LLM-powered refinement.

🖼️

Image Support

Copied images are stored as thumbnails in SQLite. Scroll through your visual clipboard history alongside text.

📌

Pin & Organize

Pin frequently-used clips so they stay at the top. Auto-classified tabs (Code / Links / Other) keep everything tidy.

🔮

Frameless Glass UI

Transparent, always-on-top window with a blur backdrop. Disappears when you switch focus — stays out of your way.

⌨️

Global Shortcut

Cmd+Shift+V summons BuffBrain from anywhere. No clicking through menus — it's always one keystroke away.

🎨

Three Themes

Dark, Light, or System — follow your macOS preference. Every detail is crafted for both modes.

🔐

Secure by Default

API keys are stored in the macOS Keychain. No config files with plaintext secrets. No telemetry. No tracking.

The brain behind the clipboard.

Every clip you copy is instantly embedded into a 384-dimensional vector by a quantized MiniLM-L6-v2 ONNX model — running entirely on your machine.

📋 Clip text 🔤 WordPiece Tokenizer 🧠 MiniLM-L6 ONNX 📊 Mean Pooling 📐 L2 Normalize 💾 384-dim vector
Model
all-MiniLM-L6-v2
quantized, ~22MB
Runtime
ONNX Runtime (ort)
No Python, no PyTorch
Search Method
Cosine Similarity
Ranked by relevance
Speed
Sub-10ms / clip
Apple Silicon M-series
"Searching for 'fruit' finds clips containing 'apple', 'banana', and 'orange' — even though none of them contain the word 'fruit'. That's semantic search."

How we stack up.

No clipboard manager has on-device semantic search. BuffBrain is the first — and it's free.

Feature BuffBrain Alfred Raycast Maccy Ditto PastePilot
Semantic Search On-device
AI Classification 30+ rules + optional LLM ⚠️ GPT cloud
Local AI Only Zero cloud by default
Price Free (MIT) £49–£69 Free + $10/mo Free Free Free
Cross-Platform macOS, Linux, Win Mac only Mac only Mac only Windows Mac only
Image Clips Thumbnails ⚠️ Premium
Global Shortcut
Frameless UI Glass, always-on-top
Privacy / No Telemetry Zero tracking ⚠️ Some ⚠️ Some Clean Clean ⚠️ Unknown
Open Source MIT GPL GPL
Platform macOS, Linux, Windows macOS macOS macOS Windows macOS

Built with modern tools.

Rust backend, React frontend, Tauri 2 framework — all tied together with a tiny ONNX model that fits inside your app bundle.

Desktop Framework
Tauri 2
Frontend
React 19 + TypeScript
Backend
Rust (tokio + serde)
Storage
SQLite (rusqlite)
AI Runtime
ONNX Runtime (ort)
AI Model
MiniLM-L6-v2 (22MB)
Secrets
macOS Keychain
Clipboard
Tauri Plugin v2

Try it yourself.

Clone the repo, build from source, and have a semantic-search-powered clipboard running in minutes. No account. No API key. Just your Mac.

Clone on GitHub Download Release
git clone https://github.com/eulogik/buffbrain.git
cd buffbrain
npm install
npm run tauri dev