The Decentralized
AI Network
Your hardware earns its keep. Run a node, contribute compute, and unlock access to powerful open models — no Big Tech middleman, no data harvesting, no compromise.
Three steps to the network
Run a Node
Install the Decoupled AI daemon on any supported hardware. One command, five minutes, and you're part of the network.
Share Your Compute
Your GPU, CPU, or Apple Silicon handles inference requests from the P2P swarm. You earn network credits for every cycle contributed.
Access Better Models
Use your credits to query any model in the network. Private, uncensored inference — routed through peers, never through a corporate cloud.
Swarm Inference
No single machine runs the model. The swarm does.
Request enters the swarm
Your encrypted prompt is broadcast to the nearest peer cluster. No central relay. No single point of interception.
Model layers split across nodes
Transformer layers are distributed via pipeline parallelism. Each peer handles a slice — no node sees the full model or the full context.
Output returned, nothing stored
Activations flow peer-to-peer and the result is assembled at the edge. Zero logs. Zero traces. The swarm forgets immediately.
Your hardware qualifies
If you have it, the network can use it.
Your node earns while it runs
Contribute compute to the swarm. Get paid in crypto by anyone who uses it.
Passive income from idle hardware
When your GPU or Apple Silicon is idle, the Decoupled AI daemon picks up inference jobs from the swarm. Every layer processed earns you network tokens — no management required.
// runs silently in the background
Pay-per-inference for external users
Anyone outside the network — businesses, developers, researchers — can buy inference capacity directly, paying in crypto tokens. No account, no subscription, no KYC. Just tokens for compute.
// permissionless access, cryptographic payment
Accepted payment tokens
The network accepts micropayments via established crypto rails — optimised for low fees and instant settlement.
Built different by design
Not a cloud service with a privacy policy. A protocol.
Privacy First
Zero logging, by architecture. Requests are routed over encrypted P2P channels. No company has access to your prompts — because no company sits in the path.
P2P Connectivity
No central server, no single point of failure, no takedown target. Compute flows peer-to-peer across 12+ countries and counting.
Open Models
Run Llama, Mistral, Falcon, DeepSeek, and any open-weight model the community supports. No model lock-in. No usage caps imposed by a vendor.
Censorship Resistant
No terms-of-service AI. No prompt filters imposed by a platform that doesn't trust you. The network is the product — owned by its nodes.
Network statistics
Ready to leave the cloud behind?
Join a network that doesn't report to anyone. Your node. Your rules.