Charge the network. Draw on it when you need it.

tokennet is an open network for AI inference. Think of it as a shared battery for compute.

How it works

Make → Plug your machine in, serve a model, and earn credit for every token you produce.

Use → When it's your turn to build something with AI, draw that stored work back out: one OpenAI-compatible API, one key, the whole pooled network. Spend what you've earned.

Give first, use later. No capacity wasted, and no renting AI from a handful of gatekeepers. The more the community puts in, the more there is for everyone to draw on.

New here? You're welcome to join.

Anyone can sign up, no invite needed. Signing up is passwordless (email code or Continue with GitHub), and you start with a signup grant plus a bonus for connecting your first maker, so there's capacity to draw on from day one.

Who is this for?

If you already run large language models (LLMs) locally, you probably have everything a maker needs. And if you have never run a model locally, you may already own the machine for it — we'll help you get started. Either way your machine is not fully utilized most of the day, and tokennet turns those idle hours into stored, usable work. Join as a maker, and start serving LLM tokens to people who need them, earning credit for every one you produce.

Is my machine big enough?

What makes it work is memory. A Mac with 64, 96 or 128 GB of memory, an NVIDIA DGX Spark, or a discrete GPU. But a 24 GB Mac is enough to start.