Agent-ready by design
The Android layer for your AI agents
Let coding agents, QA agents and your own AI workflows launch, drive, test and validate Android apps on Genymotion virtual devices. Scale in the cloud with Genymotion SaaS and PaaS, or keep everything on your machine with Genymotion Desktop.
Local or cloud virtual devices
Open-source AI Skills, MIT licence
Agent session
you Launch a Pixel on Android 15 in the cloud, install my build and check the login screen.
agent Using the Genymotion skill
$ gmsaas instances start 4c1b… login-check --max-run-duration 30
ONLINE localhost:40249
$ adb install app-debug.apk
Success
$ adb exec-out screencap -p > login.png
agent Login screen renders on Android 15. 1 issue: the Sign in button sits under the keyboard on small screens. Fix proposed in LoginFragment.kt.
16:00
LOGIN
AI SCANNING
Compatible with
Claude Code
Cursor
GitHub Copilot
Gemini
Codex
Any agent that runs shell commands
Trusted by
Your agents can write Android code. Can they run it at scale?
Generated in seconds. Tested in seconds.
A coding agent can generate a new login screen in seconds.
Then it hits a wall: no device to boot and install the APK on, no screen to check.
VIRTUAL DEVICE FLEET
ON-DEMAND CLOUD INSTANCES
SCALABLE
ACTIVE INSTANCES — 16 / 32 RUNNING
HEALTHY
ai-agent · genymotion session
$ genymotion connect --api cloud
> device: Pixel_9_API_35 · ready
$ ai-agent run --task test_login
> launching app: com.demo.app …
> screenshot captured ✓
> assert: login_button visible ✓
> tap(login_button) … success ✓
> ai: fill email + password
> assert: dashboard_loaded ✓
█
GENYMOTION CLOUD · CONNECTED
LIVE
No way to verify their own work
Agents cannot provision Android infrastructure by themselves. Without a device, "it compiles" is the best they can promise.
Physical devices don't scale to agents
Physical phones are shared, slow to reset and limited in number. Agents need dozens of clean devices at once for just a few minutes.
Uncontrolled access won't pass security review
Agents touch test builds, credentials and user data. Without isolation and control over where data goes, AI projects stall at the security gate.
Agent-ready by design
Genymotion was built to be automated long before AI agents existed. Every device can be created, driven and destroyed from a command line, which is exactly how agents work.
Scriptable end to end
Every action is available through a CLI (gmsaas for the cloud, gmtool for Desktop), standard ADB and an HTTP API. Agents already speak shell and ADB: nothing proprietary to learn.
Skills that teach agents Genymotion
Our open-source Skills give Claude Code, Cursor, GitHub Copilot and Gemini step-by-step workflows to start devices, install APKs, run tests and stream a device. One command to install.
A real Android, fully controllable
Toggle root, emulate GPS, battery, network quality, biometrics, calls and SMS, inject camera media, edit device identifiers. Agents can reproduce the edge cases that break apps in production.
Disposable and parallel
Spin up a clean device per task, run many in parallel, shut them down when the job is done.
Runs where your data must live
Our SaaS, your own AWS, Google Cloud, Azure or Alibaba account, your on-premises servers, or your laptop. One product family, same tooling.
Full observability
Stream any cloud device to a browser or embed it in a web page with our open-source JavaScript SDK. Reviewers see exactly what the agent did.
Up and running in three steps
How it works?
1
Add the skills to your agent
One command. Setup guides are provided for Claude Code, Cursor, GitHub Copilot and Gemini in Android Studio.
npx skills add Genymobile/genymotion-skills
You can also access the full source code and documentation on our official Github repository: https://github.com/Genymobile/genymotion-skills
2
Connect Genymotion
Install gmsaas for Genymotion SaaS or use gmtool with Genymotion Desktop. For Genymotion PaaS, your cloud provider’s CLI does the job.
3
Ask in plain language
The agent routes each request to the cloud or to your local device.
"What devices are available?"
"Launch a Samsung Galaxy S23"
"Install my APK"
"Run the integration tests"
"Stop my instance"
Questions teams ask before connecting to their agents
Which AI agents work with Genymotion?
Any agent that can run a shell command. We provide ready-made skills and setup guides for Claude Code, Cursor, GitHub Copilot and Gemini in Android Studio, and the same commands work from custom agents and CI jobs.
Can agents see what is on the Android screen?
Yes. Through ADB, agents take screenshots and screen recordings, read logs and dump the UI hierarchy to find buttons and fields.
Can I use Genymotion with a local, offline AI model?
Yes. Pair Genymotion Desktop with a model served by Ollama or LM Studio, for example through Android Studio Agent Mode. After activation, Genymotion Desktop runs offline for 15 days, longer on a Premium plan.
Where does my data go?
You choose. With Genymotion Desktop, everything stays on your workstation. With Genymotion PaaS, devices run in your own cloud account. Genymotion Self-Hosted runs on your servers, and Genymotion SaaS runs in the Genymotion-managed cloud.
How many devices can my agents run at the same time?
On Genymotion SaaS, contact us to raise your simultaneous-device limit. With Genymotion PaaS, capacity follows your cloud account’s quotas.
How do I keep agent usage under control?
Set a maximum run duration on every device, use quotas per group, and pull per-user usage reports.
Can I run Genymotion Desktop on a CI server?
No. Genymotion Desktop is designed for physical workstations. For CI/CD and headless agents, use Genymotion SaaS or Genymotion PaaS.
Contact us
Need more details? Feel free to contact our team with the following form.
