# Neurogrid Community Cloud

> A community-powered inference marketplace: people with underutilized GPUs supply compute, people who need to run AI models get affordable, distributed inference. Neurogrid is the marketplace connecting the two.

Neurogrid connects GPU owners with idle hardware to people who need AI inference, without either side owning or renting traditional data-center infrastructure. GPU owners contribute idle capacity and earn from the workloads that run on it; AI users reach distributed compute through an OpenAI-compatible API instead of owning the infrastructure themselves. The motto is "powered by the people, owned by the people."

## Key pages

- [Homepage](https://neurogrid.cc/): product overview, how the marketplace works, use cases for pointing existing tools at Neurogrid, the labs behind the models currently served, and how to join as a provider or user.
- [Manifest](https://neurogrid.cc/manifest): the platform's rules and values — what is and isn't allowed on the network (no intercepting or logging inference traffic that passes through your hardware, no misrepresenting contributed hardware, no malicious use of served models, nothing illegal in either party's jurisdiction).
- [Privacy Policy](https://neurogrid.cc/privacy): what Neurogrid collects (account, payment references, host hardware and approximate location, usage metering, page analytics), what it does not (prompt and response content), and that deployments run on independent hosts' machines.
- [Terms of Service](https://neurogrid.cc/terms): rules for clients and hosts, payouts every 15 days from a $10 balance through Stripe, and the ban on hosts inspecting or logging the workloads they run.

- [nrgrd-tui](https://neurogrid.cc/nrgrd-tui): Neurogrid's terminal coding agent — what it does, how to install it (`uv tool install nrgrd`, also pipx and pip), screenshots, and links to its GitHub repository and PyPI package. MIT licensed.

## How it works

1. Contribute — connect an underutilized GPU to Neurogrid.
2. Match — the marketplace routes compatible workloads to available hardware.
3. Compute — AI workloads run on distributed community infrastructure.
4. Earn — GPU providers earn from the compute their hardware contributes.

## Integration

Any client that speaks an OpenAI-compatible API can point at a Neurogrid deployment (base URL + API key) — chat clients, IDE plugins, agent frameworks, or scripts. Neurogrid also has its own terminal client, nrgrd-tui: install with `uv tool install nrgrd` (https://pypi.org/project/nrgrd/, source at https://github.com/NeuroGrid-AI-exchange/neurogrid-tui, MIT licensed).

## Community

- GitHub: https://github.com/NeuroGrid-AI-exchange
- Discord: https://neurogrid.cc/discord