RobotActions — AI-Driven Mobile + Web Test Automation

[Robot Actions — real-device testing for humans and AI agents]

Test mobile and web apps from anywhere.

Services · Pricing · Use from an AI agent · Agent Skills · File an issue

Robot Actions is a cloud-connected testing platform that lets you control real Android and iOS devices through a web browser, automate test flows with an AI assistant, and generate runnable test scripts in your framework of choice (WebdriverIO, Python, Robot Framework, and more).

For sales or enterprise questions: [email protected].

See it in action

Real iOS and Android devices, streamed live to your browser — no emulators, no local setup. A few of the things you can do:

AI Agent — explore an app and find bugs

Chat with the built-in AI Agent right inside a device session — ask it to find a defect or suggest improvements, and it drives the device on its own: screenshots, reads the screen, taps and swipes, and reports back findings and recommendations.

Real-device control — play, draw, anything

Native-speed touch on a real iPad: playing a game and sketching in Freeform, straight from the browser.

Real-time mouse & keyboard control

Native-speed pointer and keyboard input on both iOS and Android — type and click straight from your machine with no lag.

Web Inspector for mobile Safari

Full DevTools — Network, Elements, Console, Application — attached to Safari running on a real iOS device.

Chrome DevTools on Android (CDP)

Full Chrome DevTools over the DevTools Protocol — Elements, Console, Network, Application — inspecting a live page on a real Android device.

Capture & mock network traffic

Watch live HTTP(S) flows from the device, then add rules to mock, abort, or rewrite matching requests — set the status, headers, and response body on the fly.

Mock GPS location

Drop a pin anywhere — search a city or set latitude/longitude — and the device reports that location to any app, so you can test geo-aware flows without leaving your desk.

Light & dark themes

The web console follows your preference — switch between light and dark on the fly.

More feature walkthroughs are on the way.

Robot Actions on the MCP Registry

The hosted MCP server is published at https://registry.modelcontextprotocol.io as io.github.krishtoautomate/remote-device-server and auto-syncs to:

• GitHub MCP Registry (github.com/mcp) — shows up in VS Code under @mcp
• Smithery (smithery.ai) — install via the toolbox button
• glama.ai/mcp — community aggregator

Manifest source: server.json in this repo mirrors the published version.

Agent Skills

Connecting an agent to the MCP server gives it several hundred tools. It does not tell it how to use them — and left to itself an agent will tap coordinates read off a downscaled screenshot, type before the field has focus, and report success from a screen it never checked.

skills/ holds open-source Agent Skills that supply that missing procedural knowledge: which tool, in what order, and the mistakes that quietly turn a passing test into one that tested nothing.

Install into any agent — Claude Code, Cursor, GitHub Copilot, Codex, Cline, Zed and 50+ more read the same universal format:

npx skills add https://robotactions.com

Run without flags and it asks which skills and which agents you want. To choose up front, --list shows what is available without installing, and --skill <name> takes just one.

Claude Code users can install it as a plugin instead, which also registers the MCP server:

/plugin marketplace add krishtoautomate/robotactions/plugin install robotactions@robotactions

Not a Claude Code user? Each skill is a plain folder with a SKILL.md — copy what you want. skills.json is a machine-readable index. Browse them at robotactions.com/skills.

What this repo is for

This repo contains no application source code — Robot Actions is a hosted product, there is nothing to install or build. The repo exists for four current purposes plus one that's coming:

1. Public issue intake. File a bug, request a feature, or ask a question via the Issues tab. No paid account required — drive-by feedback is welcome.
2. Release notes. Every shipped change is published as a Release with user-facing notes. That's the running changelog.
3. Agent Skills. skills/ — open-source guidance that teaches an AI agent how to actually use the MCP server well. Also installable as a Claude Code plugin marketplace from this repo root.
4. Test framework examples (coming soon). Once the product is more widely used, this repo will host runnable example scripts in the test frameworks Robot Actions can generate — WebdriverIO, Python (appium-python-client style), Robot Framework, and more. These will live as small standalone projects under examples/ so you can copy-paste a starting point. None of the application source will ever land here — the examples are public starter templates, not the product.
5. Public roadmap & discussions — being scoped.

When to file an Issue here vs. when to email us

Security disclosures

Do not file security issues here. See SECURITY.md for the responsible-disclosure process.

Releases

Every shipped change is published as a Release with user-facing notes. Browse releases →

License

This README, the issue templates, the skills/ Agent Skills, and any future examples/ test scripts are MIT-licensed. The Robot Actions product itself is a hosted commercial service — not open source, and no application source code lives in this repo.