Open-source AI compute clustering
CMesh
A decentralized-ready protocol and toolkit for turning trusted machines into a private AI compute cluster.
managermac armlinux gpuwindowsstorage
StatusAlpha testnet
Current workloadDistributed compute benchmarks
Next milestoneFirst model inference job
LicenseOpen source
What problem it solves
Small teams often have idle laptops, workstations, and rented machines, but no simple way to combine them into one controlled compute pool. CMesh starts with private clusters: invite workers, set resource limits, run jobs, and measure whether more machines actually increase usable compute.
First release value
Prove connection, resource reporting, benchmark execution, and job result collection across real machines before moving into model inference.
Current capabilities
- Cluster manager with local persistence
- Worker desktop app for macOS, Windows, and Linux
- Resource limits for CPU, memory, storage, GPU, and VRAM
- Compute jobs and cluster benchmark aggregation
- Operator dashboard for workers, jobs, and benchmarks
Roadmap
- Private cluster manager with invite-based worker joins
- Desktop worker app for resource limits, status, logs, and disconnect
- Benchmark jobs that prove aggregate compute grows with workers
- Model workload runner for small inference tasks
- Storage sharing layer for model artifacts and cache coordination
Built for contributors
The project is intentionally shaped like an open-source infrastructure project: readable architecture docs, focused issues, release artifacts, and a clear path from local testing to real private clusters.