The Ethereum Foundation and the Open Anonymity Project launched zkAPI on Ethereum mainnet on October 1, 2026, enabling users to pay for metered APIs without disclosing the payer’s identity. The initial focus is AI inference, where prompts can be sensitive and conventional billing records connect each request to a user. Users deposit credits such as ETH or USDC into an Ethereum vault contract in one transaction. The resulting balance becomes a private note that only its holder can spend. Software on the user’s device generates a zero-knowledge proof (cryptographic evidence without revealing the underlying data) that the note is funded and unused. Each payment also publishes a nullifier (a one-way identifier used to prevent double spending) without exposing the note behind it. A local zkAPI client supports standard OpenAI and Ollama APIs, allowing existing applications and chat clients to operate through localhost. Servers verify payment proofs, issue short-lived API keys kept only in device memory, and deduct from the private balance when those keys expire. The payment server can confirm that a valid payment exists without learning who made it, while the AI provider can receive prompts without seeing the payer. The same system can support RPC queries, image generation, VPNs and machine-to-machine services. The implementation follows the ZK API usage credits design published on Ethereum Research by Davide Crapis and Vitalik Buterin. It uses Groth16 proofs on the BN254 curve, Poseidon hashing for commitments and nullifiers, and a 32-level Merkle tree. The live Ethereum mainnet vault holds USDC credits, while withdrawals remain available onchain if the zkAPI server stops operating. The project does not provide network anonymity, and shared prompts can still serve as fingerprints. It remains under active development, while confidential GPU computing for content privacy is still emerging.