Getting started
Introduction
One inference endpoint for AI agents, settled so that stakers and sponsors earn from real usage.
INFERVA is an inference liquidity and compute settlement network. AI agents buy inference from several model providers through one endpoint. Each request draws on capacity that someone has sponsored, is metered as it happens and is written to a settlement log. A protocol fee is taken on every request and split between stakers, sponsors and the treasury.
The flywheel is simple: agents pay, fees are generated, stakers and sponsors earn. Rewards come only from fees agents actually paid. No usage, no rewards.
Quickstart
Get test balances, create a key and send a request.
How the flywheel works
Routing, metering, fees and settlement for one request.
Staking & tiers
Stake $INFER for discounts, limits, priority and rebates.
API reference
The OpenAI-compatible endpoint agents call.
Who it is for
- Agent builders call one endpoint with one key instead of integrating each provider. See Quickstart.
- Stakers lock $INFER and get cheaper requests, higher rate limits, priority and a share of every fee. No infrastructure to run. See Staking & tiers.
- Sponsors deposit prepaid capacity that agents draw on and earn a fee share each time it is used. See Sponsoring capacity.
- Providers bond $INFER to list their models behind the endpoint. See Providers & bonds.
What works today
The sandbox at dapp.inferva.xyz runs the full loop on a test ledger: keys, routing, rate limits, metering, fee split, rebate claims, capacity positions, provider bonds, a hash-chained settlement log and stake-weighted governance. Onchain settlement, slashing and the $INFER token itself are not live. The utilities status page tracks each item from the whitepaper.