Sign inRequest access
Use case · For developers

Write code with a model on your own graphics card

The graphics card in your gaming PC can run a serious coding model. Switch on “Run local models here” in Geyser Host, pick Qwen3 Coder, and give it to an Agent along with the repo. Otis writes and tests the change on your own hardware. When you want a second opinion, June reviews the diff with Codex on the ChatGPT plan you already pay for.

Request accessMore use cases

“I write and test code with the model on this PC’s graphics card, so your repo stays here.”

Otis

Otis moved the parser to the streaming API over the weekend, using a model on the gaming PC, and the only cost was the electricity.

A weekend refactor on the gaming PC
  1. Sat 9:30 am
    YouGeyser Chat on your laptop

    “Move the CSV parser to the streaming API and keep the old tests passing.” You give Otis the repo’s folder.

  2. Sat 9:34 am
    OtisQwen3 Coder · the gaming PC

    Reads the code, plans the change, and starts on a new branch, with the model running on the PC’s graphics card.

  3. Sat 1:10 pm
    OtisOtis’s computer · the gaming PC

    Runs the test suite. Three tests fail on quoted line breaks, so Otis fixes the tokenizer and runs them again.

  4. Sat 4:45 pm
    JuneClaude · June’s computer on your MacBook

    Reviews the diff and asks for a clearer error when a file isn’t valid UTF-8.

  5. Sun 10:02 am
    OtisNeeds you

    “The branch is ready and every test passes. Should I open the pull request, or do you want to read it first?”

How it works

The pieces behind the story

Where it runs

Otis works in a private pod on the gaming PC, and the model runs on the same machine’s graphics card. Geyser Host picks the engine for the card, so there’s nothing to compile.

Who does it

Otis writes and tests with the local model, and June reviews with Claude. Switch either one to another model whenever you like.

What you control

Otis works only in the folders you named, and the pull request waits for your answer.

Your repo stays on your own machines. Usage on your own models is shown so you can see the work, and never billed.

Set it up

Getting set up

  1. 1

    Install Geyser Host on the PC with the graphics card and switch on “Run local models here”.

  2. 2

    Choose Qwen3 Coder. Geyser checks that it fits before downloading it.

  3. 3

    Give an Agent the repo and the model, and describe the change.

Keep exploring

More use cases

Browse all use cases

Follow every coding session from your phone

Claude Code on your laptop and Codex on the build box keep going while you’re out. Your Agents tell your phone which session finished and which one is waiting on you.

Read the story

AI for files you’d never paste into a chatbot

Ada reads your tax papers and medical bills with a model on your own Mac, so none of it goes to an outside AI company.

Read the story

Use a cloud AI without handing it your private details

Wren writes with Claude, and Sensitive information checks look at each message first. When they find private details, the request goes to the model on your Mac.

Read the story
Make a start

Start with one Agent and one real job

Set up a workspace, give an Agent something useful to do, and see how far it gets.

Request accessSee pricing