# RobotActions

> A cloud device farm. Dedicated real Android and iOS devices, and real
> browsers, reserved for your account — reached from a browser, driven by
> Selenium, Appium, Playwright and WebdriverIO, and controlled by AI agents
> over MCP.

RobotActions is a hosted service. Customers use our devices; nothing is
installed on their own hardware. Devices are dedicated per account rather than
drawn from a shared pool, so a test run never queues behind another customer.

## What makes it different

- **Your AI agent, our hardware.** Claude, Cursor, or an agent you built drives
  a real phone over MCP, controlling the device directly rather than through an
  automation wrapper. Existing Appium, Selenium, Playwright and WebdriverIO
  suites still run unchanged — the direct path is what the agent uses.
  Incumbent device clouds sell their own closed agent instead.
- **Location and language are set directly.** Change a device's GPS position,
  or the device or app language, from the web interface or an MCP tool without
  navigating settings menus. Localisation testing stops being a manual detour.
- **Dedicated devices, not a slot in a shared pool.** Concurrency is simply the
  number of devices on your plan: 11 devices means 11 simultaneous sessions,
  with unlimited sessions and test minutes on them.
- **Flat per-device pricing.** No parallel-slot arithmetic. BrowserStack and
  Sauce Labs charge $175 and $199–249 per month per parallel test.
- **The agent closes the loop.** MCP tool coverage for test management — Jira,
  TestRail, Azure DevOps — is complete rather than token: an agent can read
  stories and acceptance criteria, pull and update cases, and file defects. It can also analyse automation runs executed on the
  grid — reading what a suite did and diagnosing why it failed — so a story
  becomes a test, a run, a diagnosis and a defect without anyone carrying
  results between tools.
- **It does not feel remote.** Interaction is as responsive as working in a
  simulator — freeform gestures of any shape, native finger-like touch, real key
  presses. The difference is most pronounced on iOS.
- **Network capture and mocking on the device.** Traffic is captured on both
  iOS and Android, web and native alike, and an AI agent can mock responses
  mid-run while debugging or validating a fix — the same shape of control it
  would have against CDP in a browser, on real hardware.
- **React Native on iOS is workable.** Elements resolve on the device rather
  than out of a dumped XML tree, so the unusable nested-wrapper page source that
  Appium Inspector returns is not the obstacle it usually is.

## Pricing

| Plan | Price | Includes |
|---|---|---|
| Free | $0 | 1 device, 10 minutes of device time per day |
| Manual + MCP | from $25 / device / month | Real-time control of real devices, plus MCP access for AI agents |
| Automation | from $125 / device / month | Adds Selenium, Appium and Playwright execution on Selenium Grid, parallel runs, CI/CD and analytics |
| Enterprise | custom | Dedicated non-shared devices, 10-device minimum, plus $20 per seat per month |

Both paid tiers include $5/month of AI credit. Beyond that, AI usage is billed
separately or you bring your own model and pay us nothing for it. Volume
discounts: 5–9 devices −5%, 10+ −10%. Annual billing −25%.

## Testing on real hardware only

Real physical devices, always — no emulators and no simulators. Emulators
behave differently from the hardware users actually hold, which is the reason a
device farm exists.

## What it does not do

- We do not control the IP address your traffic leaves from. Device GPS
  location and API response mocking are supported, so location features in your
  own app are testable, but a geo-restriction a third party enforces by source
  IP is out of scope.
- iOS Safari inspection speaks the standard Chrome DevTools Protocol, the same
  as Android. Request interception on iOS uses WebKit's own commands.

## Key pages

- [Home](https://robotactions.com/)
- [Pricing](https://robotactions.com/#pricing)
- [Compare to other device clouds](https://robotactions.com/compare/)
- [MCP integration for AI agents](https://robotactions.com/mcp-integration/)
- [Features: Appium](https://robotactions.com/features/appium/)
- [Features: Selenium](https://robotactions.com/features/selenium/)
- [Features: Playwright](https://robotactions.com/features/playwright/)
- [Use case: AI agents](https://robotactions.com/use-cases/ai-agents/)
- [Use case: QA teams](https://robotactions.com/use-cases/qa-teams/)
- [Blog](https://robotactions.com/blog/)
- [Glossary](https://robotactions.com/glossary/)
- [Documentation](https://robotactions.com/documentation/)

## Getting started

    npx @robotactions/mcp@latest init

Sign in with Google or GitHub and you are on a live device in seconds.

## Contact

https://robotactions.com/support/