Distributed inference · self-hosted

The AI datacenter
your organization
already owns.

Nodea turns the workstations already sitting in your office — laptops, desktops, idle GPUs — into a private, self-hosted network for AI inference. No new hardware. No prompts leaving your network.

gRPC device mesh Desktop node app OpenAI-compatible API

The problem

Every completion is either a cloud invoice, or capacity you already paid for.

Status quo

Renting inference

  • Billed per token, indefinitely
  • Prompts and data leave your network
  • Capacity you don't own or control
  • Generic shared hardware
With Nodea

Idle workstations

  • Hardware already purchased and depreciating
  • Inference runs on-premises, on your machines
  • Fleet you already manage and secure
  • Spare memory sitting unused all day

The platform

A network where every laptop is an inference node, and your org is the datacenter operator.

NODEA

How it works

From idle machine to inference node, in four steps.

01

Install

The Nodea desktop app runs on macOS, Windows or Linux — on hardware the org already owns.

02

Register

The workstation opens a gRPC stream and reports its free memory as a device identity.

03

Assign

The backend matches a model to that memory footprint and pushes a load-model command.

04

Serve

The model runs locally on the workstation; generated tokens stream back to the requester in real time.

Under the hood

One wire protocol. Three platforms.

Desktop App Tauri · Rust + React bi-directional gRPC stream Backend Orchestrator Python · asyncio · PostgreSQL routes request, streams tokens WorkstationGemma-2-2B WorkstationLlama-3.2-1B WorkstationDeepSeek-R1 one .proto contract → Python, Rust, Swift

Model placement

The network fits the model to the machine — automatically.

Llama-3.2-1B

2 GB free memory

Lightweight, fast — runs on almost anything in the fleet.

Gemma-2-2B

4 GB free memory

Balanced performance and capability for everyday completions.

DeepSeek-R1-Distill-8B

8 GB free memory

Reserved for the beefier workstations — advanced reasoning.

Adoption

Point your existing stack at it. Nothing else changes.

from openai import OpenAI client = OpenAI( api_key="<your-api-key>", base_url="https://your-org.getnodea.com/openai/v1" ) client.chat.completions.create(...) # same SDK, same schema — inference now runs on your own fleet
OpenAI-compatible API Platform REST API API keys JWT desktop auth

Visibility

You can see the fleet you're running.

Live device roster 48nodes

Every registered workstation, with real-time connection status.

Per-device usage 1.2M tok / day

Tokens served today and over time, broken down by machine.

Org-wide analytics 24/7

Usage charted over time across the whole organization.

Every organization already owns an idle AI datacenter — in its employees' laptops.

Nodea turns it into a private, cost-efficient, OpenAI-API-compatible inference cloud you already control.