mcpField
R
Ramus Android

Ramus Android

Remote
ramus.devDeveloper Tools4Updated 3w ago

About Ramus Android

Ramus Android gives coding agents a disposable cloud Android emulator for testing APKs and pull-request builds. Agents can inspect and operate the device, while people can watch or take over in a browser.

Key features

  1. 01Start sessions from demo apps or pull requests
  2. 02Inspect UI trees and screenshots
  3. 03Tap, type, press, and wait
  4. 04Read device logs
  5. 05Watch or take over in a browser

Use cases

  1. 01Test an APK build with an agent
  2. 02Check a pull-request build on Android
  3. 03Inspect app behavior using UI snapshots

MCP support

The hosted Streamable HTTP MCP server exposes `android_*` tools for starting sessions, inspecting and operating the emulator, and reading logs. It can be connected by MCP clients using a Ramus API key.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://api.ramus.dev/api/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

All endpoints
Streamable HTTPhttps://api.ramus.dev/api/mcpAuth required

Tools

Details

From the project README

Ramus android-use

Android-use for coding agents, powered by Ramus. Give Claude Code, Codex, Cursor, Gemini CLI or any agent with a shell a disposable cloud Android emulator it can see, tap, type on, screenshot and read logs from, then watch or take over in your browser.

This plugin connects to the Ramus hosted service. It needs a Ramus account (free during alpha) or a 60-minute CLI trial.

Install

Claude Code, as a plugin (the android-use skill plus the Ramus MCP server; you are asked for an optional API key):

/plugin marketplace add ramus-dev/android-use
/plugin install ramus@ramus

Cursor, Codex, GitHub Copilot, Kiro and other Agent Plugins clients: install this repository as a plugin (plugin.json at the root).

Any agent that supports skills, via the skills CLI:

npx skills add ramus-dev/android-use

Claude Code, by hand: copy skills/android-use/SKILL.md to .claude/skills/android-use/SKILL.md.

Gemini CLI, as an extension (bundles the skill and the Ramus MCP server):

gemini extensions install https://github.com/ramus-dev/android-use

Any MCP client (Cline, Claude Desktop, Cursor, Windsurf, VS Code), as a remote server. Streamable HTTP, no local process:

{
  "mcpServers": {
    "ramus": {
      "url": "https://api.ramus.dev/api/mcp",
      "headers": { "Authorization": "Bearer <RAMUS_API_KEY>" }
    }
  }
}

Create the key at https://ramus.dev/settings. The server exposes android_* tools (start a session from a demo app or pull request, snapshot the UI tree, tap, type, press, wait, screenshot, logs, share); list them with tools/list. Local APK upload and the dev loop need the CLI below.

The trial needs no account:

npx ramus-cli trial
npx ramus-cli session start --apk app-debug.apk --wait

ChatGPT and Codex (OpenAI plugin)

The same repository is the OpenAI plugin package: plugin.json carries the directory listing under extensions.com.openai, mcp.json points at the remote MCP server (people connect by signing in to Ramus; no API key), and the skills declare where they run in skills/*/agents/openai.yaml. The OpenAI package ships only android-mcp (MCP tools, ChatGPT and Codex): android-use drives the CLI through the shell, which OpenAI's skill scan flags, so scripts/package-openai.py leaves it out.

Build the upload for the OpenAI plugin portal ("With MCP" path):

python3 scripts/package-openai.py   # validates, then writes dist/ramus-openai-plugin.zip

The script checks the listing against OpenAI's limits and leaves out the files for other clients.

How it works

The device is a real Android emulator running stock Google images on the Ramus fleet, streamed over WebRTC. The skill teaches the agent the observe, act, verify loop: read a snapshot of the UI tree, act on a referenced element, then check the result before moving on.

Links

Android only. Free during alpha. For testing apps you build or are authorized to test.

SKILL.md is generated from the Ramus product contracts; the canonical copy is served at https://ramus.dev/skill.md.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.