The Ethereum Foundation has launched zkAPI, a privacy-focused payment system that lets people pay for AI models and other metered online services without directly linking their wallet, account identity, and usage history. Built with the Open Anonymity Project, the experimental tool is now live on Ethereum mainnet.
zkAPI addresses a billing problem that has become common across AI products: an API key is often connected to a user account, payment method, and a record of requests made over time. That structure can give service providers a detailed view of who is using a model, how frequently they use it, and what they ask it to do. zkAPI is designed to separate payment authorization from those usage records through zero-knowledge proofs.
The Ethereum Foundation described the system in a Thursday post written by Vittorio Rivabella. The project implements an earlier proposal, “ZK API usage credits,” published on the Ethereum Research forum on February 11 by Davide Crapis, who leads the Foundation’s dAI team, and Ethereum co-founder Vitalik Buterin.
Private balances authorize API spending
The payment flow begins when a user deposits ETH, USDC, or another supported token into a vault contract on Ethereum. Instead of exposing that deposit as the balance used for a later API request, the contract records it as a private note.
When the user wants access to an AI model or another paid API, software running on their device generates a zero-knowledge proof. The proof demonstrates that the user controls a funded note with enough value to cover the intended spending limit, while withholding which specific deposit is being used.
A zkAPI server checks that proof and then issues a temporary API key with a defined spending cap. The user can send requests directly to the AI provider using that credential. When the key expires, the server charges the consumed amount against the user’s private balance.
The design gives providers a way to confirm that a customer can pay without requiring them to receive the wallet address that funded the request. It could be useful for services that charge by token use, image generation, compute time, bandwidth, or individual API calls.
Each payment also includes a cryptographic serial number known as a nullifier. Nullifiers are published to prevent the same private balance from being spent twice. If a user attempts to reuse a note, the duplicate nullifier would reveal the double-spend attempt without disclosing the user’s other payments or deposits.
Designed for existing AI applications
zkAPI has been designed to fit into software that already uses common AI interfaces. According to its documentation, the local client exposes standard OpenAI and Ollama-compatible APIs on the user’s machine.
That setup means an application can connect to zkAPI through localhost rather than requiring developers to redesign their software around a new remote interface. A chatbot, AI agent, or local developer tool could potentially direct its existing API calls through the zkAPI client, which handles proof generation and payment authorization in the background.
The initial applications outlined by the project extend beyond AI chat. They include AI agents, blockchain remote procedure call queries, image and video generation, VPN bandwidth, and machine-to-machine payments between AI agents.
The last category reflects an emerging need in autonomous software systems. An AI agent that needs to buy data, call a model, obtain compute capacity, or access a blockchain endpoint could use limited payment credentials rather than receive unrestricted access to a user’s main wallet. Temporary caps may also reduce the damage from a compromised credential, since the API key would carry only the budget allocated to it.
The GitHub repository describes zkAPI as experimental, indicating that the software remains at an early stage and may require further testing before it is used for high-value or sensitive workflows.
Privacy protections stop short of full anonymity
zkAPI obscures the onchain connection between a deposit and an API payment, but it does not make the user anonymous at every layer of the request. The project documentation explicitly states that it does not provide network anonymity.
A gateway or service provider could potentially correlate activity from a stable IP address across sessions, even when the payment proof does not reveal the user’s wallet. Users seeking stronger network privacy can route traffic through Tor, according to the documentation. Tor relays internet traffic through several servers, making it harder for a destination to identify the original connection.
The content of prompts can also expose identity. A user who includes personal details, recognizable writing patterns, workplace information, or a continuing conversation history may create links between otherwise separate sessions. zkAPI can shield the payment relationship, but it cannot prevent a person from revealing information in the data they send to an AI provider.
Those boundaries matter for services handling sensitive prompts. Privacy tools can limit data collection across payment and account systems, while the model provider may still retain requests under its own policies or identify users through technical and behavioral signals.
Part of Ethereum Foundation’s AI agent work
zkAPI comes from the Ethereum Foundation’s dAI team, which Crapis formed in September 2025 to work on decentralized AI-related infrastructure. The group also developed ERC-8004, a standard for AI agent identity that launched on Ethereum mainnet in January.
ERC-8004 focuses on how AI agents can establish an onchain identity, while zkAPI addresses how those agents or their users could privately pay for services. Together, the projects point toward infrastructure aimed at making AI systems more capable of operating independently while retaining controls over identity, authorization, and spending.
For ordinary users, the immediate value of zkAPI will depend on whether AI model providers, RPC services, and other paid API operators choose to support it. The protocol’s technical design allows services to sell metered access without tying every request to a conventional account and wallet trail, but adoption will determine whether that capability becomes a practical alternative to today’s API billing systems.
Curious how privacy shapes crypto identity? Dive deeper into anonymity and compliance in this detailed KYC and decentralization guide.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
