Ethereum

Ethereum Foundation Launches zkAPI for Private AI Payments

Built with Open Anonymity Project, zkAPI separates API requests from their funding wallets using private credits, zero-knowledge proofs and temporary keys.

Ethereum Foundation Launches zkAPI for Private AI Payments
zkAPI brings private prepaid API access to Ethereum, separating funding records from AI requests through zero-knowledge proofs and temporary spending limits. Stock illustration of an Ethereum symbol and connected computers. Illustration: GuerrillaBuzz / Unsplash

Key Notes

  • zkAPI is live on Ethereum mainnet, using zero-knowledge proofs to separate API usage from the wallet deposit that funded it.
  • Short-lived API keys cap spending and settle actual consumption, while the current browser software documents native ETH funding.
  • AI inference is the first use case, with RPC and agent payments proposed, but providers can still identify users through prompts or network data.

The Ethereum Foundation and Open Anonymity Project have introduced zkAPI, a payment system that lets users buy access to AI models and other APIs without linking individual requests to the wallet that funded them. The Foundation announced the Ethereum mainnet launch on October 1.

The project uses zero-knowledge proofs to authorize prepaid usage. Its aim is to separate a service’s billing relationship from the queries it processes, reducing the payment trail that can connect otherwise separate AI sessions to one person.

Private Credits Replace an Identifiable Billing Account

Users fund a vault on Ethereum and receive a private record of their balance, known as a note. Their device generates a proof that sufficient credits are available and have not already been spent, allowing a server to approve access without identifying the original deposit.

In the runtime-key mode, the server issues a short-lived API key with a spending limit. Prompts then travel directly from the user’s device to the AI provider. When the key expires, a signed usage receipt records the bill, and the private balance is charged for actual consumption rather than the full reserved budget.

The architecture also offers a proxy mode that relays requests through the zkAPI server. That option is simpler to operate, but the relay can see the traffic it forwards.

ETH Funding Is Documented in the Current Software

The launch announcement describes credits funded with ETH, USDC and other assets, and links a mainnet vault holding USDC. The current billing documentation, however, specifies native ETH for the browser software and its associated services. Those descriptions should not be read as a claim that every deployed vault accepts every asset.

For native ETH deposits, the browser shows an approximate dollar value that changes with Ether’s price. The documentation explicitly distinguishes that reference value from a stable-dollar balance or a conversion into USDC.

The GitHub repository provides a browser software kit and a local client compatible with OpenAI-style APIs. Users can try wallet-funded access through OA Chat or connect compatible applications to a client running on their own device. The repository labels the protocol experimental.

AI Inference Comes First, With Broader Uses Proposed

The launch implements an earlier research proposal by Davide Crapis and Vitalik Buterin. Their design explored how prepaid credits could support many requests without an identity-linked account or a separate blockchain payment for each call.

Potential applications extend to blockchain RPC queries, image and video generation, VPN bandwidth and payments between AI agents. These are possible uses of the payment model; the announcement does not establish live integrations across all those services.

The project adds to interest in privacy infrastructure for automated software, following Scroll’s proposed AI-agent pivot. zkAPI focuses specifically on separating API access from its funding source.

Private Payments Do Not Hide Prompt Contents

The Open Anonymity Project’s privacy analysis explains a central limitation: a remote model still receives the text sent to it. Personal details, distinctive writing or reused documents can allow a provider to recognize a user even when billing identifiers have been removed.

Network metadata and request timing can also support correlation. Keeping the funding trail separate therefore addresses one layer of privacy; hiding an IP address and protecting the contents of an AI conversation require additional measures.

Disclaimer: CoinScreamer is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and market insights on digital assets and related technologies. NuvexMedia LLC invests in and collaborates with companies across the digital asset, blockchain, and technology sectors. These relationships do not influence CoinScreamer’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2025 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.

Ethereum, News