Confidential Compute Open Network (COCOON) – Decentralized AI Inference on TON

COCOON enables running AI models in trusted execution environments, while earning TON cryptocurrency for compute services.

GPU owners earn TON by serving models
App developers plug into low-cost, secure and verifiable AI compute
Users enjoy AI seamlessly, with full privacy and confidentiality

This repository contains all the necessary tools and documentation to both serve and access models via COCOON.

Quick Links

For GPU Providers:

GPU providers contribute their confidential-computing GPU servers to run COCOON workloads, earning TON by securely powering AI inference for developers and their users.

Download: Latest worker release – Ready-to-run TDX image and setup scripts
Setup Guide: Instructions are included in the release archive (preview here)

For Developers:

Developers plug COCOON’s secure, verifiable AI inference into their apps and backends, so they can safely serve powerful AI features to their own users. In exchance for these inference services, they reward GPU providers with TON.

Reproducible Build

Anyone can verify the worker distribution by rebuilding from source. Note that this step is not needed to run your own workers.

# 1. Build the VM image (reproducible)
./scripts/build-image prod

# 2. Generate distribution
./scripts/prepare-worker-dist ../cocoon-worker-dist

# 3. Verify the TDX image matches the published release
cd ../cocoon-worker-dist
sha256sum images/prod/{OVMF.fd,image.vmlinuz,image.initrd,image.cmdline}
# Compare with the published checksums

The same goes for model images:

# 1. This will generate a model tar file with the full model name, which includes hash and commit.
./scripts/build-model Qwen/Qwen3-0.6B
# Compare with the published model name

Confidential Compute Open Network (COCOON) – Decentralized AI Inference on TON

Quick Links

Reproducible Build

Further Reading