Cocoon

• 0

Cocoon is designed for anyone who needs powerful AI models without handing sensitive data to centralized clouds. Requests are routed through a proxy and executed on worker nodes running in trusted execution environments, so server owners cannot see prompts, responses, or intermediate computations. This architecture lets businesses experiment with generative AI, analytics, or custom LLMs while maintaining strict privacy guarantees that are independently attestable. GPU owners can join the network by connecting compatible TEE enabled servers, automatically receiving TON payouts for each processed request according to on-chain smart contracts. Developers integrate Cocoon via a client that sends encrypted jobs and verifies cryptographic proofs that the right model was used, turning the network into a plug in backend for privacy first AI features in apps and bots. By decentralizing compute and aligning incentives, Cocoon aims to lower AI costs, increase resilience, and give users more control over how and where their data is processed.