Ash Server & Pocket Ash
A sovereign, local-first AI ecosystem. Run private RAG, execute tools, and link your devices together into a load-balanced, firewall-traversing P2P compute grid.
A Sovereign Local AI Network
P2P Compute Grid
Link your machines (desktops, laptops, NAS, or cloud VPS) together to share inference load. The worker nodes initiate outward WebSocket tunnels to the Master, bypassing NATs and firewalls with zero port-forwarding required. Stream tokens in real-time cluster-wide.
Sovereign Local RAG
Ingest documents, PDFs, and text files locally. Our C# RAG service automatically chunks, embeds, and indexes them into a local SQLite vector database. Perform lightning-fast semantic search entirely offline without sending data to third-party APIs.
Pocket Ash Mobile Familiar
Carry your AI companion anywhere. Run optimized models on-device using LiteRT, or bind Pocket Ash securely to your home Ash Server grid via Tailscale. Biometric locking ensures your AI's memories and databases remain encrypted and private.
Running the Compute Grid
Setting up a cluster node takes seconds. You can start the server in Master mode, generate a one-time pairing token, and connect worker nodes from any machine on your Tailscale mesh network.
./ash-server
# Join the grid as a worker node from another machine (e.g. your NAS or VPS)
./ash-server --worker --master http://100.78.202.51:18799 --token GRID_TOKEN
100% Open Source & Auditable
We believe in absolute transparency. All codebases for the Ash ecosystem are open source, allowing you to audit the privacy, security, and local mechanics yourself.