Ethereum Foundation launches zkAPI on mainnet for private prepaid API payments
The system uses vault deposits and zero-knowledge proofs to unlink billing from metered AI/API usage, but it does not hide prompts or network metadata.
The Ethereum Foundation says zkAPI is now live on Ethereum mainnet, shipping a prepaid payment layer for AI and other metered API services that aims to keep billing identities unlinkable from usage. The release turns a February proposal for zero-knowledge “API usage credits” into a working implementation, while explicitly limiting privacy to payment unlinkability rather than content or network anonymity.
The Ethereum Foundation said zkAPI has gone live on Ethereum mainnet, positioning it as a production implementation of a zero-knowledge-based payment design for prepaid access to AI and other metered APIs. The Foundation described zkAPI as built with the Open Anonymity Project, with the core promise being that users can pay for API usage without revealing their billing identities.
The launch is framed as a direct follow-through on a February proposal by Ethereum Foundation researcher Davide Crapis and Ethereum co-founder Vitalik Buterin for “ZK-based API usage credits.” In the Foundation’s framing, the Oct. 2 mainnet release is not a new research note. It is the earlier concept turned into a working payment flow that can be used for real requests.
Alongside the mainnet announcement, the project released tooling intended to make the system usable by developers and end users, including a local client, a software development kit (SDK), and a browser-based AI chat implementation. The announcement did not include launch metrics such as user counts, deposit totals, or settlement volume.
Vault Deposits, ZK Credit Proofs, and Short-Lived Keys: The Payment Flow and Its Privacy Limits
zkAPI’s mechanics are built around separating “who paid” from “who used,” without trying to hide what was asked or where the request came from. Users deposit funds into an Ethereum vault, then generate zero-knowledge proofs to show they have enough prepaid credit to cover an API request without revealing which onchain deposits belong to them.
Vittorio Rivabella, AI coordinator at the Ethereum Foundation’s dAI team, described the system as issuing short-lived API keys with predefined spending limits. That design choice matters because it maps onto how metered APIs are sold today, where providers enforce quotas and billing through keys, while the settlement layer can be abstracted away from the provider’s core infrastructure.
The privacy boundary is explicit. Prompts are sent directly to the AI provider, while usage is settled separately through the payment layer, which means the provider still sees the request content and the network metadata associated with it. The Foundation also noted that users can still potentially be linked across sessions through IP addresses, timing, or information contained in their requests.
In other words, zkAPI is best understood as billing unlinkability for prepaid usage, not a private AI chat stack. It can keep a deposit from being trivially tied to a specific request, but it does not claim to hide the prompt, the provider relationship, or the network-level trail that often does the real linking.
Adoption Signals Traders Can Actually Track After the Launch
The near-term market question is whether zkAPI becomes a used payment primitive or stays a proof-of-concept with a mainnet label. The announcement does not name integrated AI providers or metered API services beyond the browser-based AI chat implementation, and it does not publish adoption metrics that would let traders translate the launch into measurable onchain activity.
The cleanest confirmation path is operational, not narrative. If contributors publish contract addresses for the vault and settlement components, traders can track vault deposits, unique depositors, and settlement volume over time, and then compare that activity against integration announcements.
Integrations are the second leg. Specific provider support beyond the initial browser-based AI chat implementation would clarify whether zkAPI is being treated as a general-purpose billing rail for metered APIs, or as a narrow demo for AI usage.
The third signal is whether the client and SDK evolve to address the stated linkability risks, either by expanding privacy beyond billing unlinkability or by giving explicit guidance on recommended network privacy tooling. Finally, fee mechanics remain unresolved from the launch excerpt, including who pays gas or fees and whether usage settlement creates sustained onchain activity rather than sporadic deposits.
My Read: A Real Shipping Milestone, but the Market Signal Depends on Integrations and Measurable Usage
The launch is being read as “Ethereum shipped private AI,” and that framing doesn’t survive the fine print. What shipped is a payments primitive that unlinks billing from usage at the deposit level, while leaving prompts and network metadata in the clear for providers, which makes it closer to prepaid metering infrastructure than a privacy layer for AI interactions.
The threshold that matters is whether zkAPI gets named integrations and monitorable onchain usage once contract addresses and fee mechanics are public. If those two pieces land and settlement volume starts to look persistent, the setup starts to look structural rather than announcement-driven.
