playtesting

Claude playtests your Unity game: Editor probes, scripted Play Mode, host + client with latency, honest pass/fail report. A Claude Code skill built on Unity's Pipeline CLI.

A Codex skill for Roblox game design interviews, verified Studio prototyping, and playtest-driven iteration.

Researching autonomous game playtesting and QA using Meta AI models, with ViZDoom as the experimental environment.

Playtest agent for web games on Solari: builds the game from git in a sandbox, plays it in cloud Chrome, reports over MCP.

Autonomous game playtesting agent: plays a game, hunts bugs with generic oracles + LLM, writes reproducible reports. Includes seeded-bug eval harness and MCP server.

Agent skill: tune game feel with numbers - feel spec in tiles/seconds, fixed-tick jump sim, level reach check, latency from slow-mo video, blind A/B sign test. EN + 中文

A tiny time-loop mystery in Python, playtested by an LLM agent.

Every game-dev loop for AI coding agents (Claude Code, Codex, Cursor, Copilot, Gemini CLI): design, build, bot playtests on your real build, balance, iterate, human logs, gated itch/Steam/web shipping.

Playtest Lens: a playtest video analyst for game developers, built on Meta's Muse models (Meta Global AI Developer Hackathon 2026)

An edit-settle-verify loop for AI agents working in Unity and Godot, plus flaky-vs-real playtests and bot-vs-human calibration.

Autonomous playtesting agent for Unreal Engine 5: Muse Spark turns a plain-English design doc into checks, plays the level, and reports only bugs a replay proves.

A blind, in-character playtesting agent for games: plays through frames, keyboard and mouse, writes a first-person diary. For Claude Code and OpenAI Codex.

Runtime-agnostic project foundation for game teams using humans and coding agents. Validation, releases, playtests, living docs, and multi-agent workflows.

Game design plugin for AI agents: one-pagers, flaw reviews and ranked improvements with every finding cited, in a versioned JSON findings contract. Balance maths CLI included.

The local playtest desk, backlog and handoff you share with your coding agents: agents file checks with the command that takes you to the spot, your game teleports you there, you answer Passed or Still broken. Works with Claude Code, Codex, opencode, Antigravity and Muse Code.

Python toolkit for AI playtesting: parallel agents play real games headless (Xvfb + xdotool) via game profiles and an optional state/event probe, then reports are merged into grouped findings. Also a screenshot-driven game-automation framework.

Play Magic: The Gathering games between AI agents at speed, with a full transcript of every decision and the reasoning behind it. Forge enforces the rules, a Java bridge routes only the calls that need judgment, Python seats the players and writes it all down.

TTRPG narrative instrument — voice characters, stage scenes, and playtest tabletop RPG fiction. A Loa construct. Pairs with construct-gygax; works standalone.