.png){kind=logo}
The Ethereum Foundation and Open Anonymity Project launched zkAPI on Ethereum mainnet, introducing a private payment system for AI models and other metered APIs. Users deposit assets such as ETH or USDC into an Ethereum vault before authorizing spending through zero-knowledge proofs.
The setup separates payment credentials from API authentication and reduces the direct link between individual requests and a payer’s billing identity.
The system implements the ‘ZK API Usage Credits’ design that Ethereum co-founder Vitalik Buterin and Ethereum Foundation dAI lead Davide Crapis published in February. Open Anonymity helped turn that proposal into the client, server, proof system, and contracts now operating on mainnet.
Under the model, users first fund an Ethereum vault. Their balance then exists as a private note. When they need API access, local software generates a zero-knowledge proof showing that sufficient funds exist without identifying the original deposit.
After verifying the proof, the system creates a short-lived API key with a defined spending limit. Users can then send requests directly to the AI provider. Once the key expires, the payment system settles the metered usage against the private balance.
The payment server therefore handles proof of funds without seeing prompt content. Meanwhile, the AI provider receives prompts and responses but does not receive the billing identity behind the temporary key. One authorization can also cover a session instead of requiring an on-chain payment for every request.
The model extends beyond individual AI prompts. Ethereum Foundation documentation lists blockchain RPC requests, image generation, video generation, VPN bandwidth, and machine-to-machine services among possible uses for zkAPI.
That architecture also creates a payment route for AI agents that need to purchase services automatically. Instead of opening conventional accounts with every provider, software agents can use funded credits and temporary credentials for individual tasks.
The system also aims to reduce integration work for developers. Its local client exposes standard OpenAI and Ollama interfaces through localhost. Existing applications can therefore point to the privacy layer without requiring developers to rebuild the surrounding AI software from the ground up.
Ethereum’s role sits at the settlement layer of the process. Deposits, withdrawals, and balance closures interact with the blockchain, while the system verifies routine spending proofs off-chain. This structure avoids sending every API request through an on-chain transaction.
Read More: Will Ethereum Hit USD 4,000 Before the End of 2026?
zkAPI does not provide complete anonymity. The Ethereum Foundation says an AI provider can still see request content and network information, including an IP address. Stable network identifiers may allow gateways to correlate activity across sessions.
Prompt content can create another link. Personal details, writing style, reused conversation history, and project documents may allow an inference provider to associate separate sessions. The project therefore separates payment privacy from network anonymity and content privacy.
The software also remains experimental. Wider use will depend on integration by developers, support from API providers, user demand, and how the system performs as usage expands. The current release nevertheless moves the February research proposal into a working Ethereum mainnet implementation.
The launch arrives as Ethereum continues its broader 2026 engineering program. The Ethereum Foundation organized protocol work around scaling, user experience, and Layer-1 improvements, including plans to move the gas limit toward and beyond 100 million.
Glamsterdam also remains part of that scaling roadmap. The upgrade combines Amsterdam on the execution layer with Gloas on the consensus layer. Ethereum scheduled its Sepolia activation for October 6, while Hoodi and mainnet activation dates remain undecided.
Ethereum’s zkAPI brings the earlier ZK API Usage Credits proposal onto mainnet by separating payment identity from API access. Users fund private credits, generate zero-knowledge proofs, and receive temporary keys. The system supports AI and metered services, although network metadata and prompt content can still link activity.
