Based Launches Based AI Gateway on Hyperliquid, Brings Pay-Per-Use AI Inference to x402 Micropayments

Based has quietly opened a new front in the race to make artificial intelligence a utility you pay for by the second. The company today announced Based AI, a commoditised AI inference gateway that runs on top of Hyperliquid and settles usage via the x402 payment rails, with privacy-focused GPU clouds operated in partnership with Privy. The gateway, reachable at ai.based.one, is designed so that software agents can pay for the “right to think” in tiny, instant micro-payments rather than through subscriptions or prepaid credit bundles.

The idea is simple but consequential: if autonomous agents and AI-driven services are going to make commerce decisions on behalf of humans, or directly as independent economic actors, they will need a frictionless way to buy compute and models. Based AI turns inference into a pay-as-you-use commodity. Developers building agents on Hyperliquid can plug into the gateway, keep a small stablecoin balance in a self-custodial wallet, and have payments deducted automatically per inference request. Based frames this as the missing primitive for what it calls “agentic commerce,” where countless small programmatic transactions must settle with instant finality and negligible gas friction.

Technically, the gateway leans on the promise that Hyperliquid offers a gasless, high-throughput environment with instant settlement characteristics that make high-frequency decision-making possible. That’s precisely the use case x402, a micropayment protocol that leverages payment-proof flows modeled on HTTP 402-style interactions, is intended to address: API endpoints, compute and data can become billable resources that a machine can purchase on the fly. Based’s team argues that billing systems built like traditional SaaS simply won’t cut it when agents are firing off thousands of tiny requests per minute.

Privacy is another selling point. Based says its Privacy Cloud runs open-weight models on self-hosted GPU clusters with a strict “no storage, no logs, no caches” policy for privacy-sensitive workloads. The pitch is that developers and agents can run LLMs without surrendering raw prompts and user data into third-party logs or reuse pools, an important distinction as more commercial use of LLMs collides with regulatory scrutiny and competitive privacy claims. Based positions this approach as an alternative to the familiar “opt-out” settings that most AI providers hide deep in their terms.

Built for Agentic Commerce

For teams already building on Hyperliquid, the onboarding is intentionally lightweight: hold more than $1 worth of USDC or USDH in a self-custodial wallet, start calling the gateway, and the balance is debited automatically. The gateway intends to be model-agnostic, offering a single integration point that can route or switch between models without forcing developers to rebuild integrations as model vendors or pricing changes. That simplicity is part product design and part strategic positioning: if agents are to transact continuously, they’ll favor low-friction, switchable plumbing.

Based’s timing fits into a broader ecosystem shift. Recent coverage and funding signals around Hyperliquid-powered apps have elevated the platform as a likely host for next-generation, high-frequency financial and commerce primitives; meanwhile, multiple projects and exchanges are experimenting with x402-style micropayments so that machine-to-machine and agent-driven commerce can become practical at scale. By bundling instant settlement, per-request billing, and privacy-first inference into a single gateway, Based is staking a claim that intelligence itself will become a consumable commodity in the near future.

Whether autonomous agents will soon be paying for their own compute in tiny slices is still an open question, but Based AI is an early production attempt to make that future operable. The gateway is live at ai.based.one and, for now, requires no KYC or credit-card plumbing: a small on-chain balance and a handful of API calls are all that stand between a developer and consumption-based AI inference. For anyone building agents or exploring programmatic commerce, it’s a development worth testing sooner rather than later.