Darkbloom is a private inference network built on idle Macs: instead of sending every prompt to a centralized cloud, models run on a distributed network of Apple Silicon machines. This repository, from Layr Labs, contains the native provider app, the local inference runtime integrations and the project site. It has about 557 stars on GitHub and was actively updated in October 2026.
The provider software runs models locally using Apple’s MLX framework. A Mac can serve an OpenAI-compatible API for its owner’s own use, or connect to the Darkbloom network so its spare compute becomes part of a shared inference pool. The emphasis is privacy and distribution: inference capacity comes from many independent machines rather than one provider’s data center. The repository is a Swift codebase for the provider and device-local APIs, and it also retains a snapshot of the consumer console. For developers, the OpenAI-compatible surface means existing tools and SDKs can talk to a Darkbloom provider with minimal changes, whether that provider is their own Mac or the wider network.
The project comes from Layr Labs, the team behind EigenLayer, and applies a similar decentralized-network idea to inference: many individually owned machines contributing capacity to a shared service, rather than a single company operating every server in the pool.
The provider software in this repository is free to download and run on your own Mac. Inference consumed through the Darkbloom network may be priced separately, so check darkbloom.dev for the current network terms.
Darkbloom is for Mac owners interested in contributing idle compute to an inference network, and for developers exploring private, distributed LLM inference instead of centralized APIs.