Privacy-preserving inference
Route confidential prompts through onion circuits so no single node sees both source and content.
Sending sensitive prompts — legal drafts, medical notes, proprietary code — to a centralized provider means one operator can read both who you are and what you asked. Metadata alone can be revealing, and you have to trust an invoice and a privacy policy.
- Optional 3-hop onion routing splits the path so no single node observes both the request source and the prompt content.
- The OpenAI-compatible gateway still handles auth, rate limits, and audit logging without decrypting your payload end-to-end.
- Differential privacy on any federated round you participate in keeps gradient updates from leaking training data.
- Because it is MIT-licensed and self-hostable, you can run the whole path on infrastructure you control.
Try it for privacy-preserving inference.
Read the deep dive, then run the quickstart.
Other use cases
DePIN GPU compute that earns
Put idle DePIN or spot GPUs to work as an OpenAI-compatible inference node — dual SOL revenue from inference requests and federated training rounds.
Federated model training
Coordinate training across nodes that never share raw data, with FedAvg, FedProx, FedAdam, and SCAFFOLD.
Machine-payable inference for agents
Give an autonomous agent an OpenAI-compatible endpoint it can pay for per call in SOL — point the existing SDK at the gateway, keep your app code unchanged.