localhost

Install an AI model on your laptop. Build something real with it. We ship you parts for the next one.

No API keys. No cloud bills. No data leaving your machine. LLMs, vision, voice — just you, Ollama, and whatever you can dream up.

~/my-project
$ ollama pull gemma4:e2b
pulling manifest... done
$ ollama run gemma4:e2b
>>> help me build a study app
I'd love to help. What subject
are you studying? I can generate
flashcards, quizzes, or explain
concepts in simpler terms.
I run entirely on your machine.
Your notes stay private.
gemma4:e2bgemma4:e4bphi-4 miniqwen3:4bllama 3.2qwen2.5-codergemma4:26bmistraldeepseek-codergemma 3ncodellamaphi-4 reasoninggemma4:e2bgemma4:e4bphi-4 miniqwen3:4bllama 3.2qwen2.5-codergemma4:26bmistraldeepseek-codergemma 3ncodellamaphi-4 reasoning
the idea

Every AI product you use sends your data to someone else's computer. Your conversations, your homework, your personal notes, all processed on servers you don't control.

What if it didn't have to be that way?

Ollama lets you run real AI models on your own laptop — language, image, voice, and more. No internet required. No API key. No cost per token. You pull a model, you run it, you build on top of it. Everything stays on 127.0.0.1.

localhost is a Hack Club You Ship, We Ship. Build an app powered by a local AI model. Ship it. We ship you hardware to build the next one.

the process
1

Install Ollama

One command. Mac, Windows, Linux. Two minutes. You now have a local AI runtime.

2

Build something

Pull a small model. Write an app that talks to it. Track your hours on Hackatime.

3

Ship it

Open source on GitHub. README with setup instructions. A demo video showing local inference.

4

Extend it

We mail you parts for your next build — mics, cameras, sensors, or a custom macropad.

suggested models

What fits your machine

LLMs are the most common starting point, but image (generation & recognition) and voice models count too — as long as they run locally on your machine.

Gemma 4 E2B any laptop
~5 GB RAM
Runs on basically anything. Vision, function calling, 128K context. Apache 2.0
Gemma 4 E4B dense
~8 GB RAM
The sweet spot. Beats Gemma 3 27B on benchmarks at a fraction of the size
Phi-4 Mini 3.8B dense
~6 GB RAM
Punches above its weight on math, reasoning, instruction following
Qwen 3 4B dense
~6 GB RAM
Hybrid thinking mode, strong multilingual, great for chat apps
Llama 3.2 3B any laptop
~6 GB RAM
Creative writing, chat companions, game NPCs
Qwen 2.5 Coder 7B 16gb+
~10 GB RAM
Best coding model at this size. Code review, generation, debugging
Gemma 4 26B 24gb+ MoE
~18 GB RAM
128 experts, only 3.8B active per token. Best quality/speed ratio if you have the RAM
the loot

You build. We ship.

Ship a project and pick from the hardware pool — parts to help you build your next one. Mics, cameras, sensors, microcontrollers, inputs. Track your coding on Hackatime so we can see the work behind the build.

Microcontroller dev board

The brain for anything physical. Wire it to buttons, screens, motors — whatever your next project needs.

Sensor pack

Motion, distance, temperature, light, touch. Feed real-world signals into your local models.

USB microphone

Plug it in, pipe audio to a local voice model. Dictation, wake words, ambient transcription — your call.

Camera module

Small, cheap, fast. Point a vision model at the real world instead of at your screen.

Programmable macropad

Custom keys bound to your models. Hotkey voice dictation, one-button image capture, anything scriptable.

Inputs & outputs

Encoders, buttons, LEDs, small displays, speakers and more!.

the fine print

What counts as shipped

Runs locally

Your app runs inference locally on consumer hardware. LLMs via Ollama, llama.cpp, or LM Studio. Image and voice models via any local runtime. No cloud inference.

Has a real interface

Web app, desktop app, or something visual. Not just piping text through a terminal. Someone non-technical should be able to use it.

Open source

Public GitHub repo. README with setup instructions. Modelfile included (if not publicly available). Multiple commits showing real work over time.

Demo video

Short recording showing the app working with the local model on your machine.

Tracked on Hackatime

Install Hackatime so your coding time is logged automatically. We use it to see the work behind the build.

Not a wrapper

Must do more than forward prompts & be a chat UI. Add context, memory, data processing, a creative concept. Make it yours. (see examples for inspiration)

inspiration

What people have built

Real projects running real models on real hardware. Every page load shows a different set, or you can browse them all.

your laptop is
the data center

Join #localhost on the Hack Club Slack. Install Ollama. Start building.

hack club you ship, we ship · built on 127.0.0.1