Repolode
ZSeven-W

ZSeven-W/dsh-game-studio

A DeepSeek Harness plugin that turns a one-sentence game idea into a playable browser game in the chat sidebar, with optional AI art and one-file HTML export.

Stars
173
Forks
22
Open issues
0
Last commit
yesterday

What is it?

DSH Game Studio is a plugin for DeepSeek Harness that builds small browser games from a plain-language description. The agent starts from a 2D top-down or 3D voxel template, reshapes it into the requested game, and shows the result in a sandboxed sidebar panel that rebuilds on every change.

Characters begin as code-drawn models. For real art, the agent generates numbered candidate images per slot through the OpenAI image API or the local Codex CLI, after showing an approval card with the service, image count and expected cost. A headless Chrome or Edge playtest runs the latest build and reports phases, score, frame rate, errors and screenshots. Every change is saved as a version that can be restored, and the finished game exports as one offline HTML file or a temporary read-only share link.

Key features

  • 01Generates a playable 2D or 3D game from one sentence using built-in templates.
  • 02Shows a live preview in the DSH sidebar that rebuilds on each change.
  • 03Creates numbered AI art candidates via OpenAI or Codex, with a spending approval step.
  • 04Runs headless playtests that report score, errors and screenshots.
  • 05Saves version history and restores code, rules and art together.
  • 06Exports a self-contained offline HTML file and offers a local-network share link.

Use cases

  • Prototyping a small casual game idea quickly inside a DeepSeek Harness session.
  • Making a simple collect-and-dodge or puzzle game with custom art swapped in.
  • Sharing a finished browser game as a single HTML file on any static host.
  • Testing a game on a phone over the local network via a QR code link.

Quick start

  1. 1

    Install the plugin into a DSH profile from npm

    dsh plugin --profile web add @zseven-w/dsh-game-studio@latest
  2. 2

    Start DSH in the web profile

    dsh web
  3. 3

    In a session in a workspace folder, ask the agent for a game in one sentence

Things to know

  • Playtests need Google Chrome or Microsoft Edge installed on the same computer.
  • AI art requires an OpenAI API key or a signed-in Codex CLI; otherwise games use code-drawn characters.
  • Requires DeepSeek Harness 0.2.1 or later and Node.js 24.11 or later.
  • The plugin is at release candidate stage (0.1.0-rc.2) and has few stars.

In collections

Topics

FAQ

What is dsh-game-studio?
A DeepSeek Harness plugin that turns a one-sentence game idea into a playable browser game in the chat sidebar, with optional AI art and one-file HTML export. DSH Game Studio is a plugin for DeepSeek Harness that builds small browser games from a plain-language description. The agent starts from a 2D top-down or 3D voxel template, reshapes it into the requested game, and shows the result in a sandboxed sidebar panel that rebuilds on every change.
Is dsh-game-studio free to use? What license does it have?
dsh-game-studio is open source under the MIT license. Check the license terms before using it in commercial products.
Is dsh-game-studio still maintained?
Actively maintained. The last commit to dsh-game-studio was on Nov 6, 2026.
What can I use dsh-game-studio for?
Prototyping a small casual game idea quickly inside a DeepSeek Harness session. Making a simple collect-and-dodge or puzzle game with custom art swapped in. Sharing a finished browser game as a single HTML file on any static host. Testing a game on a phone over the local network via a QR code link.
How do I install dsh-game-studio?
Install the plugin into a DSH profile from npm: dsh plugin --profile web add @zseven-w/dsh-game-studio@latest Start DSH in the web profile: dsh web In a session in a workspace folder, ask the agent for a game in one sentence
Where can I get dsh-game-studio?
The source code and installation instructions are on GitHub: https://github.com/ZSeven-W/dsh-game-studio

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects