OrogenVerifiable LLM inference, anchored to physical work.

OpenAI-compatible gateway routed to TEE-attested GPU operators, with validator replay, opML challenges, and burn-and-mint settlement on a Substrate-based chain.

Start here

Protocol RFCs

orogen.network

Inference customers, operators, and the chain can all check.

Every response ships a signed receipt (RFC-0001). Validators replay ≥10% of jobs on identical hardware (RFC-0006). Slashing extrinsics are on-chain with reason codes (RFC-0005). Verifiability is pallet-enforced, not a marketing claim.

How receipts work

Drop-in OpenAI compatibility — your code works unchanged.

`pip install orogen-sdk` or `npm install @orogen/sdk` and point `base_url` at the gateway. The wire shape is OpenAI Chat Completions plus four optional `useful_*` fields (nonce, tier, region, price cap). A LiteLLM provider ships the same surface inside existing stacks.

Customer guide

USD-denominated payouts via burn-and-mint, not speculation.

Customers pay USD. The gateway burns OROG at the oracle TWAP (RFC-0008) and credits the customer in non-transferable CUC. Operators serve inference and earn freshly-minted OROG — 75% operator, 15% verification, 5% treasury, 5% governance.

Tokenomics

Demand-elastic emission, pallet-enforced — no foundation mint discretion.

Per-epoch mint is bounded by a 180-day rolling burn cap with a 5%/year supply ceiling and a 0.5%/year floor. Year 1 bootstrap is 8%, Year 2 4%, Year 3+ demand-elastic. Mint outside the rule is impossible at the runtime layer.

Emission rule

What lives in these docs

These docs cover the four audiences identified in the marketing brief. If a section is short, that is intentional — the canonical specs live in the RFCs and the per-pallet code under pallet-suite/; prose here exists to orient, not to duplicate.

Audience	Job to be done	Where to start
Customers — app builders, LLM consumers	Send OpenAI-shaped requests, verify receipts, estimate per-tier price	/customers/
Operators — GPU owners across six tiers	Run a node, pass attestation, stake, route to demand	/operators/
Validators — replay node operators, stakers	Earn the 15% verification slice by replaying sampled inferences	/validators/
Researchers, holders, governance	Read the protocol spec, the RFCs, and the tokenomics	/protocol/ · /tokenomics/ · /governance/

Status, as of May 2026

Pre-build planning. Mainnet TGE targets Q2 2027. See the roadmap and the live STATUS.md.
Public testnet (Forge) opens Q1 2027. The customer SDKs and LiteLLM provider exist today as drop-in OpenAI clients pointing at a placeholder gateway — see Downloads.
Open decisions are tracked in DECISIONS.md.

If something here contradicts the marketing brief or the canonical RFC, the RFC wins and this page is the bug. Open an issue or edit on GitHub.