Open-source MCP lets Claude playtest Windows games in the background across 20+ titles
Wonderful-Ad-8579 · reddit · 2026-11-06
A developer open-sourced an MIT-licensed MCP server that injects per-game virtual keyboard, mouse, controller input and screen capture inside the game process, so an LLM agent can play in the background while you keep working—with multiple games running in parallel. Works with Claude Code, Codex or any MCP client; Windows only, and refuses to inject into games with anti-cheat.
The demo video shows a Sonnet agent replaying early Totally Accurate Battle Simulator campaign levels: it reads the enemy army off the screen, places its own, and retries with a different army after a loss.
Findings from 20+ games of playtesting:
- Screenshots dominate token cost; agents downscale/crop and use OCR to read text. Two sessions cost 1% of a weekly Sonnet limit.
- Unity, turn-based, menu-heavy games work best: TABS, Terraria, Oxygen Not Included, Phantom Rose, Halo Wars 2.
- Older engines are hard: some DirectX 9 and Unreal 4 games ignore background input; Skyrim's mouse only worked in the latest release—10 releases in 2 days.
Workflow: Claude Code planned and reviewed, Codex wrote most code, Sonnet agents playtested and logged bugs. Author uses it mainly for playtesting and mod development—the mod-writing agent can launch the game and verify its own changes.
Related event: Open-Source MCP Lets Claude Play Windows Games in the Background(2 posts)→
More from coding & agent
- Nous Research ships Hermes Agent v0.23.0 with desktop-in-browser and unified gateway — NousResearch · 2026-11-06
- Opik Launches Generative Dashboards: Plain-Language Agent Builds Custom Charts — dl_weekly · 2026-11-06
- Aggregating all agent decisions into one prioritized 'needs me' review queue — kevinkern · 2026-11-06
- AWS WAF lets developers charge AI agents for content, paid in stablecoins via Stripe — jeff_weinstein · 2026-11-06
- Ripple-65B flops in single-GPU coding test against Qwen 27B despite SOTA claims — DZhang50 · 2026-11-06
- Essay argues ontology is AI's next frontier as reasoning outpaces our definitions of reality — _AustinCalvert_ · 2026-11-06