LeashLEASH
On-device · Private · Est. 2026The private exocortex
Your mind, on your own devices.
Leash is a private assistant grounded in your own data — Apple Notes, files, memory, and your world. It runs entirely on your devices. No cloud. No leak. Powered by your personal mesh.
Private
End-to-end encrypted, and yours. Your data never leaves your devices — no server to subpoena, no model trained on you.
On-device
Every token, embedding, and fine-tune runs on your own hardware via QVAC. It works in airplane mode.
Your mesh
Your phone, laptop, and desktop become one brain. Borrow a bigger model or more compute from whichever device has it.
Inside Leash
The feature
An economy of agents.
Your devices form a market for intelligence.
When a node needs more model than it can run, it doesn’t fall back to the cloud — it borrows compute from another device on the mesh and pays per token, settled on-chain (x402-style machine payments). Providers earn; small hardware runs big models. A real machine-to-machine economy, live across your own devices.
See it on GitHub
…and more
How it works
Inference & embeddings
GGUF models run locally through QVAC — chat, RAG, and vision, all on your hardware.
Encrypted P2P mesh
Devices pair over an end-to-end-encrypted DHT — no broker, no cloud relay.
Delegated compute
A node that needs more model borrows it from another and pays per token.
On-device fine-tuning
LoRA adapters (QVAC Fabric) personalize the model without sending a byte off-device.
Built for QVAC Hackathon I — “Unleash Edge AI”. All on-device, by design.
Put your AI on a leash.
Private, on-device, yours. Download Leash for your platform.