OpenGrid

Pre-production · Apache 2.0

Independent operators.Unified by protocol.

The AI Grid brings compute from independent providers into one intelligent network. Each request is placed by capacity, reputation, and locality, and verified with a signed receipt naming the operator, model, and build that served it.

Protocol
v1.6
Licence
Apache 2.0
Access
Waitlist
coordinatorclientsmachinesclusters
who generates

Machines with owners

A Mac, a workstation, a 1MW modular site — enrolled by whoever owns it. Ownership never transfers to us. The node key is generated on the machine and never leaves it.

how you call it

Nothing to rewrite

An OpenAI-compatible surface your client already speaks. The migration is one base URL.

what comes back

Evidence, not assurances

Every successful call returns a signed record naming the exact model artifact, runtime build and operator that answered.

Have capacity?

From a Mac to a megawatt.

One protocol covers both ends. An M-series Mac on a desk takes a share of latency-tolerant work; a modular datacenter with power and spare utilisation takes production traffic. Same enrolment, same identity model, same settlement. Your hardware stays yours, and your cluster keeps its own scheduler.

Join the waitlistSee what a request looks like

Your key never leaves the box
The node identity is generated on the machine you enrol.
Signed work, signed settlement
Completed work is countersigned and posted to a double-entry ledger.

The coordinator, the node daemon and the protocol are Apache 2.0. Every claim on this page is specified in the repository, where you can verify it for yourself.

Read the source