Technology
An orchestrator, a memory, a rail.
Rivier is three things working as one. A learned orchestrator that routes every request to the model that earns it. A memory you own that grounds every route. A real settlement rail that clears the work. Each layer is built, not borrowed.
Layer 01
Rivier One — the orchestrator
Every request is routed in the moment to the model that answers it best per dollar — a cheap specialist we own (Tier I), an open model on the Tenzro network (Tier T), or a frontier model (Tier F) when the task earns it. Cost, latency and privacy are inputs to the route, not afterthoughts. The industry baseline puts 30–60% of LLM spend on overkill; routing fixes that bill without sacrificing quality.
Layer 02
The memory it routes against
Rivier holds a private, encrypted picture of who you are — your taste, your plans, the way you like things done — on Tenzro, the decentralized network for identity, memory, compute and payments. Every route grounds in this. Frontier models receive the personalisation; they never touch the raw graph behind it, and they never see anything that identifies you.
Layer 03
Identity and settlement
One Rivier Identity binds who you are to how you pay. RIVR is the accounting layer: one balance, seen in your own currency, with every confirmed action carrying a signed receipt you can audit. And it settles for real — on Canton, the same institutional-grade ledger the markets run on. Your balance isn't a number in an app. It's money that actually clears.
Specification
The platform, plain.
- MemoryEncrypted personal graph on the Tenzro network. User-owned, portable, never handed to the host assistant.
- ReasoningRivier One — learned hybrid-routing conductor over owned, network and frontier tiers. OpenAI-compatible endpoint with cost, latency and privacy ceilings per call.
- ActionScoped agent identity plus the RIVR settlement spine — book, buy and pay within user-set limits, on explicit approval.
- SettlementReal money that clears on Canton — the same institutional-grade ledger the markets settle on — not just a balance in an app.
- DistributionAn OAuth-protected MCP server. Same six verbs on Claude, ChatGPT, and the Rivier web. Connect once, used everywhere.
- ReceiptsEvery reasoning call returns a signed CortexReceipt — model id, weights hash, input/output commitments. Surfaced as a checkmark; raw receipt available on demand.
Read the whitepaper.
The full architecture, the commitments, and the model under which Rivier ships. Available to partners and operators on request.