The Morpheus Inference Marketplace is where decentralized AI becomes accessible to everyone. It connects people who need compute power (consumers) with providers who supply it, all coordinated by smart contracts on BASE. Instead of a centralized service, requests and responses flow directly between consumer and provider in a peer-to-peer system. The Morpheus Lumerin Node is the open-source software stack that wires actual prompts and inference to that marketplace. It is the missing piece between “there is a contract on chain” and “I am chatting with a model.”

​
Why decentralized inference matters

Decentralized inference breaks the dependency on centralized providers who control access, pricing, and availability of AI infrastructure. In a decentralized system:
  • Anyone can contribute resources, and anyone can use them — an open, resilient, competitive market.
  • No single points of failure. No one provider’s outage takes the network down.
  • Costs stay fair through transparent on-chain incentives, not vendor pricing power.
  • Innovation isn’t gated by the decisions of a few large players.
It’s about shifting power from closed platforms to an open network where inference is a shared, community-driven resource.

​
Design

At the centre of the marketplace are the Compute Node contracts (the Diamond marketplace on BASE). They:
  • Register providers and the models they host.
  • Match consumers with available providers.
  • Secure connections through encryption and verifiable on-chain logic.
To participate, both sides run a proxy-router — a lightweight binary (this repo) that talks to the contracts and manages session lifecycle. Once a session is open, the consumer’s prompts go directly to the provider’s model and the responses come back the same way — no middleman, no inference flowing through a Morpheus-operated server.

​
Three core ideas

Peer-to-peer after match

Once a session is open, prompts and responses go directly between consumer and provider — no Morpheus-operated server in the data path. The blockchain only sees session open / close / settle, never the inference itself.

Session-time pricing, not tokens

Pricing is per-second of session time, not per-token. Consumers stake MOR for a fixed duration; the MOR is escrowed on open. On close, unused stake returns immediately and the used stipend day-locks until the next UTC day. See Sessions: stake, close, claim.

Reputation-weighted matching

Providers with strong uptime, time-to-first-token, throughput, and success rate are matched more often by the consumer-side rating system. See rating-config.json for how the weights work.

​
What Morpheus is not

  • Not a hosted service. There is no “Morpheus, Inc.” running the inference. Providers are independent operators.
  • Not on-chain inference. The blockchain coordinates the marketplace; it does not execute prompts.
  • Not a wallet. MorpheusUI manages a key for convenience; the keys, ETH, and MOR are all standard ERC-20 / EOA primitives on BASE.
  • Not token-priced. Pricing is denominated in session time (pricePerSecond), not tokens. Long contexts and short contexts cost the same per second.

​
Two ways to use Morpheus

You can either run a node (peer-to-peer, self-custody, native to the marketplace) or use the hosted API Gateway (no node, simple API key). See Inference API overview for the hosted path.

​
Where Morpheus, NodeNeo, Everclaw, app.mor.org fit

See Ecosystem for curated, mirrored summaries with attribution.

​
Further reading (canonical, off-site)