nunu.ai

Overview

nunu.ai runs AI agents that test games and apps the way a human tester would. You upload a build (APK, IPA or a Windows ZIP), describe the test in plain English, such as "complete the tutorial and flag issues", and an agent runs it on a real device in nunu's infrastructure by watching the screen and using touch, keyboard or mouse. Every run produces a full recording with the agent's reasoning and any bugs it found.

Because the agents work from pixels, there's no engine integration to start, and an optional lightweight SDK adds error logs and app state. It runs on iOS and Android devices and, since June 2026, full Windows PCs for PC and Steam games; console support is listed as coming soon. Agents can run in parallel and can collaborate on multiplayer tests, and paid plans add scheduled test plans, MCP and API access.

Key features

• Plain-English test instructions, no scripts
• Vision-based agents on real iOS, Android and Windows devices
• Session recordings with reasoning and bug reports
• Parallel runs and multi-agent multiplayer tests
• Optional SDK for logs and app state
• MCP and API access on paid plans

Use it for

Pricing

Starter is $0 with 2,000 one-time credits, full platform access, 5 GB of build storage and session recordings. The self-serve plan starts at $200/month for 20,000 credits (up to about 40 hours of agent work) with one concurrent device, 60+ real iOS and Android devices, test plans and schedules, 100 GB of storage, SDK, MCP and API access. Enterprise is custom, and a persistent private device is a $500/month add-on.

Getting started

Sign up at nunu.ai, upload a build to Build Storage, write a Discovery test in plain English and hit Play, then review the recording under History.

Open nunu.ai

Alternatives

modl.aiAI bots that playtest your game to find bugs and performance issues.ObvraTurns recorded gameplay into reusable AI playtests with visual targets and failure evidence.