In a hurry? Pick the fastest path.
- Just want to use it (~5 min): download the desktop app from the latest release and run it — wallet, chat, and proxy-router are bundled. See the Consumer quickstart.
- Try it free — no wallet, no tokens: the desktop app bundles a local
llama.cppmodel for demos, so you can chat immediately without staking MOR. Look for the local model option after install. - Building on the marketplace? jump to the Hosted Inference API.
What’s in the box
proxy-router
Background process that listens to the BASE blockchain, manages secure consumer/provider sessions, and routes prompts and responses. The same binary serves both sides; only configuration differs.
MorpheusUI
Electron desktop GUI that consumers use to browse bids, open sessions, and chat with models.
mor-cli
Command-line client that talks to the proxy-router HTTP API.
llama.cpp + tinyllama
A throw-away local model bundled for demos so you can try the stack without paying MOR.
How the pieces talk
- Consumer opens a session by staking MOR against a provider’s bid on chain.
- The consumer proxy-router opens a TCP connection to the provider proxy-router (port 3333 by default) and forwards prompts.
- The provider proxy-router dispatches to the configured backend model (
apiUrlinmodels-config.json) and streams the response back. - The BASE blockchain holds the source of truth for providers, models, bids, and sessions. MOR pays for usage; ETH on BASE pays for gas.
Pick your role
Hosted Inference API
No node, no wallet — OpenAI-compatible API key.
Consumer quickstart
Install the desktop release, fund a wallet, open a session, chat.
Provider quickstart
Stand up a proxy-router, look up existing models on active.mor.org, bid (mint only if needed).
TEE provider fast lane
Deploy a hardened
-tee image on SecretVM in minutes.Resale provider
Resell Venice / OpenAI / Anthropic capacity on Morpheus.