So we built it.
Why we built this
There are GPUs in universities that sit idle between semesters. Research clusters that run at 30% capacity. Gaming PCs that sleep 16 hours a day. Data centres with spare racks. Personal machines with more power than their owners will ever use.
The hardware exists. It's scattered across the world — in labs, offices, homes, and facilities — waiting for something to do.
Cloud made compute accessible — and that was a massive leap forward. But most of the world's hardware still isn't connected. It sits outside those systems, unused.
What was missing was a way to bring it all together — cloud, on-prem, personal, research, edge — into one network where anyone can contribute and anyone can build.
We built idyl to connect what's already there.
What we believe
Not a product you buy from a vendor. A network you participate in. Open by default. Controlled where it matters. Owned by the communities that use it.
The network protocol is documented and public. HTTP and SSE with mTLS. No proprietary wire format. Open at every layer.
The CLI, agent, and SDKs are open source. Anyone can read the code, contribute, or build their own tooling. Transparency is not optional.
Anyone can deploy workloads. Anyone can provide compute. No approval process to get started. The network welcomes everyone.
Subnets belong to the communities that build them. The people who contribute compute, deploy workloads, and shape the network — they are the network.
The team
We build and operate the idyl network. The CLI, agent, and core libraries are open source. The platform services that run the control plane are closed source — for now.
Open source CLI, agent, and SDKs. Read the code, contribute, or build your own tooling.
github.com/idyl-labs →Architecture, API reference, getting started guides. Everything you need to build on or contribute to the network.
docs.idyl.network →We're building something that matters. If you want to work on hard problems with real impact, we want to talk.
View openings →