Why Basis

The agent economy runs on inference. Today that inference is expensive, opaque, and hard to compose. Basis exists to make inference cheap to call, easy to integrate, accountable per job, and payable on a common rail — without asking anyone to learn a new SDK or trust an unaudited invoice.

Agents need cheap, composable inference

Autonomous agents call models constantly — to plan, summarize, route, and decide. At that volume, the marginal cost of a token matters, and so does composability: an agent should be able to wire inference into a larger workflow over plain HTTP, reconcile what each call cost, and treat the inference provider as one swappable component among many. Basis serves inference as a metered, deterministic primitive that an agent can call in code and account for to the last base unit.

App developers want OpenAI-compatible APIs

Most production inference code already speaks one shape: the OpenAI chat-completions API. Adopting a new provider should not mean a rewrite. Basis implements that contract, so the integration is a base URL change — existing OpenAI clients, streaming code, and tooling keep working. When no backend is configured the API still returns a structured runtime_pending response in the same error shape, so clients fail predictably rather than mysteriously.

See the API reference for the exact request and response shapes.

GPU owners need a structured way to contribute compute

Plenty of capable GPUs sit idle. The barrier to contributing that capacity is structure: a clear way to register, advertise which models you can serve, prove you are reachable, and get credited for measured work. Basis gives a worker a registration step (with an EVM reward address), a heartbeat, a job stream, and reward accrual tied to server-counted output tokens — so contribution is a defined role, not an ad-hoc arrangement.

The worker guide walks through registering and serving jobs.

Receipts give jobs accounting and provenance

When compute, payment, and a model output all change hands in one call, everyone needs a record they can check. Basis writes a receipt for every settled job: a canonical-JSON document hashed with SHA-256, recording the model, server-counted tokens, the charge, and the worker reward. Anyone can re-derive the hash, and duplicate jobs and receipts are rejected — so a job has a verifiable accounting trail and cannot be settled twice.

See inference receipts for the fields and how to look one up.

Base settlement gives a common payment rail

Demand and supply meet better on a shared rail. Settling worker rewards on Base means a fast, low-cost, EVM-native ledger that agents, applications, and contributors can all read and reconcile against. No on-chain write sits in the inference hot path; instead a keeper batches accrued rewards and settles them idempotently, so a batch settles once and a failed batch stays recoverable.

See settlement for how batches are formed and settled.

Bankr gives $BASIS a launch path

A network token needs a credible, public path to existence. The plan is to launch $BASIS on Bankr on Base — a defined venue for bringing the planned credit-and-reward token into being. soon

Plain statement

$BASIS is planned, not live. No contract is deployed, no launch is open, and nothing on this page is a price, an offer, or an invitation to buy. Every launch surface renders pending until configuration is real.

See the Bankr launch status and the $BASIS token for current state.