Robotactions — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited Robotactions (MCP Server) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
<p align="center"> <a href="https://robotactions.com"> <img src="assets/banner.png" alt="Robot Actions — real-device testing for humans and AI agents" /> </a> </p>
<p align="center"> <strong>Test mobile and web apps from anywhere.</strong> </p>
<p align="center"> <a href="https://robotactions.com/#services">Services</a> · <a href="https://robotactions.com/#pricing">Pricing</a> · <a href="USING_MCP.md">Use from an AI agent</a> · <a href="skills/">Agent Skills</a> · <a href="https://github.com/krishtoautomate/robotactions/issues">File an issue</a> </p>
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].
<!-- mcp-name: io.github.krishtoautomate/remote-device-server -->
Real iOS and Android devices, streamed live to your browser — no emulators, no local setup. A few of the things you can do:
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.
<video src="https://github.com/user-attachments/assets/83326ac2-2c53-4b47-b7e7-625641064bc0" controls width="640"> <a href="https://github.com/user-attachments/assets/83326ac2-2c53-4b47-b7e7-625641064bc0">▶ Watch the demo</a> </video>
Native-speed touch on a real iPad: playing a game and sketching in Freeform, straight from the browser.
<video src="https://github.com/user-attachments/assets/263e71d6-c03b-4753-975e-2c1fa413b64c" controls width="280"> <a href="https://github.com/user-attachments/assets/263e71d6-c03b-4753-975e-2c1fa413b64c">▶ Watch the demo</a> </video>
Native-speed pointer and keyboard input on both iOS and Android — type and click straight from your machine with no lag.
<video src="https://github.com/user-attachments/assets/70ce3e9a-2846-47c5-af42-9c0763ce8419" controls width="640"> <a href="https://github.com/user-attachments/assets/70ce3e9a-2846-47c5-af42-9c0763ce8419">▶ Watch the demo</a> </video>
Full DevTools — Network, Elements, Console, Application — attached to Safari running on a real iOS device.
<video src="https://github.com/user-attachments/assets/dd40cff5-ea4f-4f7f-8495-18027368c54d" controls width="640"> <a href="https://github.com/user-attachments/assets/dd40cff5-ea4f-4f7f-8495-18027368c54d">▶ Watch the demo</a> </video>
Full Chrome DevTools over the DevTools Protocol — Elements, Console, Network, Application — inspecting a live page on a real Android device.
<video src="https://github.com/user-attachments/assets/da84182d-70e1-4e77-aa57-5d8a095c133e" controls width="640"> <a href="https://github.com/user-attachments/assets/da84182d-70e1-4e77-aa57-5d8a095c133e">▶ Watch the demo</a> </video>
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.
<video src="https://github.com/user-attachments/assets/d3e8b177-2983-4940-9bd5-52d945c55bf4" controls width="640"> <a href="https://github.com/user-attachments/assets/d3e8b177-2983-4940-9bd5-52d945c55bf4">▶ Watch the demo</a> </video>
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.
<video src="https://github.com/user-attachments/assets/d93d160f-6a8e-4210-9d92-f92aa30b1cbb" controls width="640"> <a href="https://github.com/user-attachments/assets/d93d160f-6a8e-4210-9d92-f92aa30b1cbb">▶ Watch the demo</a> </video>
The web console follows your preference — switch between light and dark on the fly.
<video src="https://github.com/user-attachments/assets/b5ebee16-9c1a-49db-9487-afc3603c8a83" controls width="640"> <a href="https://github.com/user-attachments/assets/b5ebee16-9c1a-49db-9487-afc3603c8a83">▶ Watch the demo</a> </video>
More feature walkthroughs are on the way.
The hosted MCP server is published at <https://registry.modelcontextprotocol.io> as io.github.krishtoautomate/remote-device-server and auto-syncs to:
github.com/mcp) — shows up in VS Code under @mcpsmithery.ai) — install via the toolbox buttonManifest source: server.json in this repo mirrors the published version.
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.comRun 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| Skill | Covers |
|---|---|
| robotactions-getting-started | Connect, authenticate, verify, device holds and the parallel limit |
| mobile-app-testing | Driving native Android and iOS apps on real hardware |
| accessibility-testing | Accessibility audits and walking the real TalkBack / VoiceOver screen reader |
| web-app-testing | The web on a real device browser or the desktop grid |
| device-state-setup | Files for upload flows, GPS, locale, dark mode, clean app state |
| network-mocking | Mocking and capturing traffic — native app requests as well as web |
| framework-integration | Using the tools with an existing Appium/Selenium/Playwright suite |
| flow-record-replay | Recording replayable regression tests with assertions |
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.
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:
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.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.| Filing here | Email us at [email protected] |
|---|---|
| Bug reproductions, feature requests, public questions | Account-specific issues (billing, credentials, dedicated deployments) |
| "Does Robot Actions support X?" | Anything that would expose your account email, device IDs, or session data publicly |
| General feedback / ideas | Sales conversations, enterprise contracts |
Do not file security issues here. See SECURITY.md for the responsible-disclosure process.
Every shipped change is published as a Release with user-facing notes. Browse releases →
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.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.