Image: GitHub
Why it mattersAn indie developer who already pays for Claude or ChatGPT gets a game-dev workspace that routes work across those models and their asset tools, with no new subscription for the app itself.
A desktop app for building games where the AI is the Claude Code or ChatGPT subscription the user already pays for, and the models run across the main agent, worker agents and reviewers, is the shape of Genex. The project is on macOS Apple Silicon, Linux x64 and a Windows x64 preview build, with 636 stars in the 16 days since it was first published on GitHub, and a .dmg, .deb, .rpm and unsigned Windows installer linked from the latest release.
What it does
The project page describes Genex as a workspace that uses the user's Claude Code or ChatGPT subscription, or local models, and mixes Opus and GPT models across three roles: a main agent, worker agents and reviewer agents. Local 3D assets come from a Blender plugin, and more asset work routes through the Genex tools layer to Meshy, Tripo and ElevenLabs. The app exports to disk or publishes to the web, and ships Unreal and Godot plugins so a developer can build on a Mac in either engine. The README names C++ games as planned.
Plugins carry the tools
Plugins give Genex's agents new abilities: an asset generator, a bridge to another engine, or a service the game talks to. The Blender integration and the Genex tools layer that ship with the app are themselves plugins. A developer scaffolds a new plugin with npm run plugin:new, checks it with npm run plugin:doctor, and loads it into the app from Plugins, Add, Load local plugin. A plugin shared on a public GitHub repository can be installed from there by anybody, and a pull request against the genex-plugins repository lists it in every Genex app's Marketplace.
Where it fits
The point of running an indie game-dev workspace against a user's existing AI subscription is that the three roles of the pipeline, the main agent writing code, workers producing assets or scenes, and reviewers checking the output, each pick their model from the account the developer already has. A game studio using Claude Opus for engine code and GPT for narrative work does not add a third per-seat bill to that stack. For hobby and single-developer projects, the whole app runs on one Claude Code or ChatGPT seat, which is the deciding cost in a side project.
Genex is early. The README says so plainly, Windows builds ship unsigned as a preview until SignPath signing is wired up, and every release lists each file's SHA-256 in a SHA256SUMS file so a user can check the download. The project is maintained by the genex-games GitHub organisation, under MIT, and the contributing guide requires macOS Apple Silicon, Git and Node 24.
The audience is specifically a developer building a game and holding an existing AI subscription. For a solo developer who has been pasting prompts between a chat window and the engine, an app that drives Unreal or Godot through the same subscription and lets Blender work land through a plugin, is a change to the shape of the day. The marketplace of community plugins is the thing to watch: a Meshy-level asset generator written by somebody else is what makes a workspace worth living in.
Source
The repository is on GitHub at genex-games/genex-desktop. The project site is at genex.games.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.