A blind, end-to-end-encrypted marketplace for LLM inference, paid in bitcoin, built on the NUTS ecosystem. Providers run Ollama and choose which models to sell; consumers run any coding agent (via Nemesis8) against those models. A central gateway authenticates both sides with NUTS tokens, matches them, relays opaque ciphertext, settles payment, and takes a fee — without ever seeing a prompt or a reply.
Charon is the ferryman: it carries traffic across, gets paid, and learns nothing about its passengers.
- Client — the consumer/provider proxies. The consumer proxy exposes a plain OpenAI-compatible API locally and does all the marketplace work behind it (identity, encryption, payment). The provider proxy runs next to Ollama.
- Service — the gateway (relay): a blind matchmaker that connects, relays ciphertext, settles payment, and aggregates reputation.
Specifications
The authoritative description of the protocol lives in spec/. Start
with spec/README.md for the reading order, or jump to:
- 00 — Overview
- 01 — Architecture
- 02 — Identity & Auth
- 03 — Wire Protocol
- 04 — Encryption
- 05 — Payments
- 06 — Provider
- 07 — Consumer & Nemesis8
- 08 — Discovery & Reputation
- 09 — Gateway
- 10 — Security & Threat Model
License
BSD 3-Clause. See LICENSE.