Pistachio: Peer-to-Peer AI Inference on the Morpheus AI Network

Pistachio is a local AI inference gateway on the Morpheus decentralized network. Pistachio connects to the Morpheus network, which offers token-staked inference sessions, open-weight models, and permissionless access.

MOR tokens are staked, not spent. You put MOR into a smart contract to open a session with a provider. When the session closes, the full amount returns to your wallet. The only direct cost is ETH gas for the on-chain stake and unstake, which on Base is fractions of a cent per session.

How is that possible? Who pays the providers? Morpheus runs on a model called the Techno Capital Machine. The protocol uses token emissions funded by staked yield to compensate compute providers. When you stake MOR to open a session, you signal demand, and the protocol routes emissions to the providers serving it. Providers get paid by the network, not by you. Your MOR is not a fee. It is capital you are tying up for the duration of the session. The real cost is the opportunity cost of that tied-up capital, not per-request spending. Capex, not opex. You deploy capital, it works, it comes back. The MOR you hold today opens a session, returns when the session closes, and opens another one tomorrow.

Providers run open-weight models and there is no platform-level content policy. Any provider can serve any model they choose. The models respond to what you ask, not to what a centralized compliance layer permits. This matters for researchers, writers, developers, and anyone whose work requires a model that responds to the question asked.

Access is mediated by a smart contract, not a company. You do not need an account, an API key, or approval from anyone. The network is also working toward trusted execution environments (TEEs) to strengthen privacy at the provider level. No single entity controls who can use the network or what they can run.

Getting started

You install Pistachio locally (currently in early alpha, live demo here) and connect your wallet. You need a Base wallet (MetaMask works), MOR tokens (available on Aerodrome or through Coinbase DEX), and a small amount of ETH on Base for gas.

When you send an inference request, Pistachio stakes MOR from your wallet into the Morpheus contract, which opens a session with a provider serving the requested model. The provider sees the stake and accepts prompts. Pistachio routes your prompt via the MOR RPC protocol. The provider runs the model and streams the response back through a standard OpenAI-compatible API. Your existing tools (Cursor, any OpenAI-compatible client) work without modification.

When the session ends, the staked MOR returns to your wallet in full. The only cost is a fraction of a cent in gas per stake and unstake on Base.

The Pistachio Pass: capacity-based NFT access

Pistachio is free to start. Install it, then claim a free DRM3Pass NFT (about a cent of gas on Base) to get 20 MOR of staking capacity. That is enough for short sessions, roughly 23 minutes of continuous inference before the session cycles.

For longer sessions, upgrade passes are available as ERC-721 NFTs on Base. The system is additive: each pass adds capacity on top of what you already have. Bronze adds 200 MOR (about 4 hours continuous). Silver adds 1,000 MOR (about 19 hours). Gold adds 5,000 MOR (about 8 days). Diamond adds 20,000 MOR (about 6 weeks).

The NFT is not a subscription. It is a capacity token. You are not paying for inference. You are paying for the ability to stake more MOR at once, which means longer sessions. Longer sessions matter for two reasons: you burn less ETH gas because you restake less often, and you have fewer gaps in availability. With a small stake, sessions are short, which means Pistachio has to close and reopen sessions frequently. Each cycle is a window where you have no active inference lane. A larger stake means longer continuous uptime, fewer interruptions, and less gas. The pass saves you money and keeps your inference available.

Available models

Morpheus providers run open-weight models. The current catalog includes Qwen 3 (235B, 32B), LLaMA 3.3 70B, DeepSeek, and others depending on which providers are online. Pistachio includes a model catalog with live health tracking so you can see what is available, response latency, and provider status.

On a centralized API, the provider decides which models you access. On Morpheus, any provider can serve any model they choose to run. If one provider is slow or unreliable, Pistachio routes to another. No single point of failure, no single point of control.

Provenance on every response

Every inference response through Pistachio carries an Ed25519 provenance receipt. The receipt documents the model, the provider, the token count, the response hash, and the session ID. Every AI output is cryptographically traceable: you can prove which model produced it, when, and through which provider.

This is the same signing infrastructure DRM3 uses across RSS feeds, DNS scans, and data collection. One protocol, every layer of the stack. The receipt format and verification process are identical whether you are verifying an article analysis or an AI inference response.