Build a free game with AI.
FreeGameStore is AI-first. Drop our SKILLS.md into your AI tool of choice, and the agent already knows the brand rules, the privacy mandates, the CLI, and the deploy flow before it writes a line of code. Ship a real, free, MIT-licensed game in an afternoon.
Even faster: connect the MCP server and build, publish, and improve games without leaving your editor — your AI writes the code (create_game → update_files), or the platform's VibeCode agent does (agent_build). One line to connect:
Pick your tool
Claude Code
Anthropic's official terminal CLI. Best fit if you live in a shell.
EditorCursor
AI-native fork of VS Code. Best fit if you want a polished editor UI.
EditorGitHub Copilot
Built into VS Code, JetBrains, and the GitHub web UI. Best fit if you want zero install friction.
TerminalAider
Open-source AI pair programming in your terminal. Best fit if you want to swap models freely.
TerminalCodex CLI
OpenAI's open-source terminal agent. Best fit if you prefer GPT-5 / o4 models.
EditorWindsurf
Codeium's AI-native editor with Cascade for multi-step tasks. Best fit if you want agentic flows in an IDE.
EditorZed
Fast Rust-based editor with multi-model AI. Best fit if you want low-latency editing + AI in one tool.
EditorContinue
Open-source AI extension for VS Code and JetBrains. Best fit if you want full local-model freedom.
EditorCline
Autonomous AI coding agent for VS Code. Best fit if you want the agent to plan and act without hand-holding.
WebChatGPT (web)
Build via the ChatGPT browser UI when you don't want a local agent. Best fit for one-shot prototyping.
Workspace layout
Keep all your game repos under one folder so your AI agent finds them automatically:
This matches the paths in SKILLS.md. For apps, use ~/dev/fas/apps/.
Or any other tool
Roo Code, Sweep, Open Hands, Smol Developer, the new Anthropic SDK agent — all of them work. The pattern is the same:
- Install the FreeGameStore CLI:
npm i -g @freegamestore/clithenfgs login. - Scaffold:
fgs init my-game --template canvas. - Tell your AI tool to read SKILLS.md. Drop the URL in the project's instruction file (CLAUDE.md, .cursorrules, .github/copilot-instructions.md, AGENTS.md, CONVENTIONS.md, system prompt, …).
- Build. The agent edits
web/src/App.tsxand supporting files using the Shell layout, brand tokens, and brand fonts. - Run
fgs check. Compliance enforced. - Run
fgs publish. Provisions repo + hosting + DNS. - Push to main. Auto-deploys in ~30 seconds.
Drop that URL in your tool's instruction file. The agent fetches it on demand and writes platform-aware code.
What "AI-first" actually means here
Three things, enforced by tooling:
- Brand and privacy are not negotiable. Your agent will try to add Google Analytics or a custom display font.
fgs checkrejects the publish. The mandates exist precisely so that AI-generated apps stay consistent. - Provisioning is one command. No copy-pasting Cloudflare tokens. No DNS dashboards. No CI YAML.
fgs publishhandles everything atomically. - Deploy is a push. The agent doesn't need a deploy script.
git pushtriggers Cloudflare Pages, the storefront refreshes within hours.
The mandates aren't to slow agents down — they're to let agents go fast without breaking the platform.
Other reading
- Manual / non-AI build flow — the same steps without an agent.
- Platform guidelines — what gets approved and what doesn't.
- FreeGameStore SDK source — CLI, compliance package, backend.
- FreeAppStore's AI guide — sister storefront, same flow.