The Ethereum Foundation and Open Anonymity Project launched zkAPI on Ethereum mainnet on October 1, 2026, enabling metered API payments without disclosing the payer’s identity. The initial focus is AI inference, where sensitive prompts and billing records can otherwise connect each request to a user. Users deposit ETH or USDC into a vault smart contract (self-executing blockchain code) in one transaction, receiving a private note that only its holder can spend. A local client creates a zero-knowledge proof showing that the note is funded and unused, while a nullifier identifies double-spending attempts without exposing the note behind the payment. The client supports standard OpenAI and Ollama APIs through localhost, allowing existing applications to continue operating. Servers verify payment proofs, issue short-lived API keys kept in device memory, and debit the private balance when those keys expire. The design separates information between the payment server, which verifies payment without identifying the user, and the AI provider, which receives prompts without seeing the payer. zkAPI is based on a design published on Ethereum Research by Davide Crapis and Vitalik Buterin, using Groth16 proofs on BN254, Poseidon hashing and a 32-level Merkle tree. The live mainnet vault holds USDC credits, and withdrawals remain available onchain if the zkAPI server disappears. The project does not provide network anonymity, and shared prompts can still create identifying fingerprints; confidential GPU computing for content privacy remains an emerging area.