Ethereum Foundation and Open Anonymity Project Deploy zkAPI on Mainnet

The Ethereum Foundation and the Open Anonymity Project have launched zkAPI on Ethereum mainnet, introducing a protocol that allows users to pay for metered API access without exposing their identities to service providers ¹. The release implements an earlier design co-authored by Davide Crapis and Ethereum co-founder Vitalik Buterin ¹.

Under current web architecture, metered API services such as AI inference require users to provide API keys tied directly to corporate accounts and payment methods ¹. zkAPI alters this flow by splitting payment settlements from usage data ¹. Users deposit assets like $ETH or $USDC into an onchain vault contract, creating an offchain balance represented as a private note ¹. To fund API calls, a local client generates Groth16 zero-knowledge proofs on the BN254 curve, proving that a funded, unspent note exists within a 32-level Merkle tree ¹. The zkAPI server verifies these spend proofs offchain and provisions short-lived, dollar-capped API keys directly to the client's memory ¹.

Cryptographic Isolation and Limitations

Once authorized, client requests travel directly to the API provider using the ephemeral key ¹. When a key expires, the provider issues a signed usage receipt, prompting the zkAPI server to deduct the exact metered amount rather than the initial spending cap ¹. Because nullifiers prevent double-spending and commitments mask the funding origin, providers observe user prompts but cannot identify the payer, while payment servers record metered charges without accessing request contents ¹. If intermediary servers go offline, depositors retain the ability to exit and recover their remaining funds directly from the onchain vault contract ¹. At the time of writing, ETH traded at $2,705.

The developers noted that zkAPI addresses billing links but does not provide complete anonymity ¹. API providers still process raw prompt contents and IP addresses, meaning users must employ network protections like Tor or confidential computing environments to prevent traffic correlation or content-based fingerprinting ¹. Beyond AI inference, the project stated that the system is designed to support private metered billing for blockchain RPC nodes, image generation, machine-to-machine agent tasks, and VPN bandwidth ¹.