Use your email or service name to access the operator dashboard.
Create a service account. You will receive a verification email, then your request will be reviewed before you can sign in.
Verifying your email address…
Peer-to-peer LLM inference. Point your apps at a local proxy — models run on your machine, or on a trusted peer when you need more compute.
Public Beta — features may change. Terms & Privacy.
mtrxAI connects independent machines into a peer-to-peer inference mesh. Each node runs an OpenAI or Ollama-compatible proxy; the lobby handles discovery and signaling only — payloads stay between peers.
localhost:11345; models run on whichever peer has them loaded.Pull the container, join a cluster or swarm, point your apps at the local proxy. Hardware sharing is always optional.
View on Docker HubAn intelligent layer between your apps and a peer-to-peer compute mesh — local-first by default, mesh fallback when you need it.
Browse models across your trusted network or deploy your own. Run locally when hardware permits, or borrow from peers when VRAM is limited.
Strict local-first policy. If the model is on your GPU, the workload stays local — fastest path, zero network overhead.
When remote inference is needed, mtrxAI picks the best provider by LAN, geography, and current load.
Clusters for managed teams with discovery, ranking, and credits. Swarms for token-less decentralized P2P rings.
Works with existing AI tools, agent stacks, and custom workflows — no application code changes.
Credits settle when consumer and provider agree on usage. Beta tokens have no monetary value.
One endpoint for Ollama, vLLM, LM Studio, and other backends — a single catalog for your apps.
Wake idle models on the mesh. mtrxAI finds a peer with enough VRAM to pull and load the weights.
Sharing windows, thermal guards, and a kill-switch to isolate your machine instantly.
For invite-only team connectivity without VPN setup, see Private clusters.
Pull mtrxAI from Docker Hub and register your device ID in the dashboard.
Create or join a private cluster, a public regional cluster, or a Swarm with a shared token.
Connect Ollama, vLLM, or other engines. mtrxAI merges them into one catalog.
Redirect AI tools to the local proxy. Sharing hardware is optional — approve each request or stay isolated.
One proxy gateway for your stack. Access teammates’ models without copying multi-gigabyte weights.
Offload heavy inference to a peer when local hardware is tight, then switch back instantly.
Share idle GPU time and earn credits for when you need external compute.
Isolated private clusters — internal models hidden from public discovery.
One way to use mtrxAI’s peer-to-peer mesh: create an invite-only cluster for your team. Models stay on the machines that load them; your apps keep talking to localhost.
Install mtrxAI on each machine, create or join a private cluster, and point your applications at the local proxy. When a model isn’t local, mtrxAI routes the request to a peer in the same cluster over a direct WebRTC channel.
Each machine uses the lobby over WebSocket for discovery and connection setup. Once peers connect, prompts and responses travel directly — the lobby is not in the data path.
Develop on a lightweight laptop while heavy models run on a rig with enough VRAM. Your IDE still targets localhost:11345.
Link multiple workstations into one cluster. Shared models stay invisible to the public mesh; LAN peers are preferred automatically.
Peers connect via outbound WebRTC (STUN/ICE hole punching). No static IPs, no router admin, no opening inbound ports in typical NAT setups.
On the workstation that will host models, create a private cluster in the dashboard. Copy the cluster name, UUID, and password.
On your laptop or additional workstations, join with the same name, UUID, and password. Each machine gets its own peer identity.
Attach LLM backends, enable “share with cluster” on providers, and connect the cluster in the dashboard. Remote models appear in your local catalog.
Most home and office routers work out of the box. Restrictive corporate NAT may require future relay support.
You control visibility, sharing, and who connects. Inference payloads on the cluster path travel peer-to-peer; the lobby coordinates discovery and signaling only.
Beta notice: Do not route regulated or highly confidential data through public clusters. See Terms & Privacy.
Each node runs a local Ollama-compatible proxy. The lobby discovers peers and relays connection metadata; inference stays between machines whenever possible.
Mix topologies concurrently — regional clusters for discovery and credits, swarms for small trust circles.
| Managed clusters | Decentralized swarms | |
|---|---|---|
| Primary use | Teams and regional communities with discovery, ranking, and credit tracking. | Small high-trust groups with minimal central coordination. |
| Joining | Public regional index or private cluster (name + UUID + password). | Shared cryptographic Swarm token. |
| Economics | Transparent credit accounting between peers. | Honor system — no ledger. |
| Transport | Lobby signaling, then direct WebRTC data channels. | libp2p gossip — fully decentralized coordination. |
| Network setup | No inbound ports; STUN/ICE hole punching for WebRTC. | libp2p hole punching with optional relay fallback. |
Private team connectivity is documented as a use case in Private clusters.
Multi-backend proxies, peer discovery, WebRTC inference, smart routing, credit accounting.
Peer reputation, dynamic pricing signals, confidential GPU attestation.
Local-first by default. You decide what to share, when, and with whom.