# agemu (npm · agentic-emulator)

Fast Android emulator and iOS Simulator control for coding agents: screenshots, UI tree, taps, logs

- Trust score: 82/100 (high trust)
- Change this week: +18
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-11-21

## Components

- npm · `agentic-emulator`: 82/100 (this document), [markdown](https://verifymcp.io/servers/piotr-graczyk-dev-agemu/agentic-emulator.md), [page](https://verifymcp.io/servers/piotr-graczyk-dev-agemu/agentic-emulator)

## Channel facts

- Registry: `npm`
- Package: `agentic-emulator`
- Version: `0.2.2`
- Transport: `stdio`

## Trust breakdown

How this component scores in each security and reliability category. Every signal is checked automatically from public evidence about the published package, including repeated runs of it in an isolated sandbox, and we only credit what we can confirm. Scores are 0–100 per category. Scoring method: https://verifymcp.io/docs/scoring (what has changed: https://verifymcp.io/docs/scoring/changelog)

Scored 2026-11-21.

- **Supply Chain Security**: 100/100
  - No malware found by supply-chain analysis.
  - No known CVEs affecting this package version or its production dependencies.
  - No install/post-install scripts declared.
  - No production dependencies, so there is no dependency health to assess.
- **Provenance & Transparency**: 100/100
  - Source repository is publicly reachable at the declared URL.
  - Cryptographically verified build provenance (signed, bound to piotr-graczyk-dev/agentic-emulator).
  - Clear OSI-approved license (MIT).
  - Actively maintained (last published 5 days ago).
  - Publishes a security disclosure policy (SECURITY.md).
- **Schema Quality & AI Usability**: 69/100
  - AI-judged instruction clarity (good).
  - Context-footprint check failed: tool/resource definitions use about 3642 tokens (~117/item across 31 items; 31 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 23/100
  - Stability observed for 7 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **Tool Coverage**: 100/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 100% of tool parameters carry a description.
- **Tool Safety**: 75/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - 0 of 3 tool(s) whose name or description implies an irreversible operation declare an MCP destructiveHint annotation; "shutdown" implies "shutdown" and declares no destructiveHint at all, which the MCP spec reads as destructive by default.
  - An AI judge read all 32 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 100/100
  - Implements a supported MCP spec version (2025-11-25); the latest is 2026-07-28.

## Install

### How do I install the agemu MCP server?

agemu runs locally as an npm package, launched with npx -y agentic-emulator mcp. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

### Claude

```bash
claude mcp add piotr-graczyk-dev-agemu -- npx -y agentic-emulator mcp
```

### Cursor

```json
{
  "mcpServers": {
    "piotr-graczyk-dev-agemu": {
      "command": "npx",
      "args": [
        "-y",
        "agentic-emulator",
        "mcp"
      ]
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "piotr-graczyk-dev-agemu": {
      "command": "npx",
      "args": [
        "-y",
        "agentic-emulator",
        "mcp"
      ]
    }
  }
}
```

### Codex

```bash
codex mcp add piotr-graczyk-dev-agemu -- npx -y agentic-emulator mcp
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "piotr-graczyk-dev-agemu": {
      "type": "local",
      "command": [
        "npx",
        "-y",
        "agentic-emulator",
        "mcp"
      ],
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add piotr-graczyk-dev-agemu --command npx --arg -y --arg agentic-emulator --arg mcp
```

### Hermes

```yaml
mcp_servers:
  piotr-graczyk-dev-agemu:
    command: "npx"
    args: ["-y", "agentic-emulator", "mcp"]
```

### Netclaw

```json
{
  "McpServers": {
    "piotr-graczyk-dev-agemu": {
      "Transport": "stdio",
      "Command": "npx",
      "Arguments": [
        "-y",
        "agentic-emulator",
        "mcp"
      ]
    }
  }
}
```

### Vellum

```bash
assistant mcp add piotr-graczyk-dev-agemu -t stdio -c npx -a -y agentic-emulator mcp
```

### Other

```json
{
  "mcpServers": {
    "piotr-graczyk-dev-agemu": {
      "command": "npx",
      "args": [
        "-y",
        "agentic-emulator",
        "mcp"
      ]
    }
  }
}
```

## Changelog

Every change recorded for this component, newest first. Days that predate change tracking, or that we cannot explain, say so: "we were watching and nothing happened" and "we were not watching" are different claims.

### 2026-11-20 (score 82, +1)

No change was recorded against any check on this day. Stability & Change Management went from 17 to 20. That category is still filling its 30-day observation window: 5 days of observed history at the previous scan, 6 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-18 (score 81, +1)

No change was recorded against any check on this day. Stability & Change Management went from 10 to 13. That category is still filling its 30-day observation window: 3 days of observed history at the previous scan, 4 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-16 (score 80, +16)

- [security improvement] Malware scan: unverified → pass

### 2026-11-15 (score 64, 0)

- [security improvement] Malware scan: unverified → pass
- [functional improvement] Stability: unverified → 0.03
- [functional] Package version: 0.2.1 → 0.2.2

### 2026-11-14 (score 64)

First indexed and scored.

## MCP tools (31)

### `devices` (~21 tokens)

List Android and iOS devices with state, profile and screen size.

### `profiles` (~30 tokens)

List screen profiles (real device sizes) usable with set_profile.

Input parameters:

- `platform` (string): android or ios

### `create` (~40 tokens)

Create a new device.

Input parameters:

- `name` (string, required): Device name
- `platform` (string, required): android or ios
- `profile` (string): Profile name

### `boot` (~163 tokens)

Boot a device to an agent-ready state; from its baseline snapshot this takes seconds and returns state booted. The first boot of a new Android device (or a baseline retake after a GPU mode, AVD configuration or agemu upgrade) prepares the baseline once, which usually takes 7-8 minutes and never more than 20 (then the boot fails): boot then returns at once with state "preparing" and the phase; poll devices every 30-60 s (or call boot again) until the state is booted. Other calls on a preparing device fail at once with an error saying so.

Input parameters:

- `cold` (boolean): Force a cold boot
- `device` (string): Device id or name (optional when exactly one device is booted)

### `shutdown` (~32 tokens)

Shut a device down.

Input parameters:

- `device` (string): Device id or name (optional when exactly one device is booted)

### `reset` (~40 tokens)

Restore a booted device to its clean baseline state (fast).

Input parameters:

- `device` (string): Device id or name (optional when exactly one device is booted)

### `park` (~72 tokens)

Park a booted device: stop it in place, so it uses no CPU and the host can reclaim its memory, keeping its apps and screen as they are. The next call on it resumes it (or use resume).

Input parameters:

- `device` (string): Device id or name (optional when exactly one device is booted)

### `resume` (~49 tokens)

Resume a parked device; returns once it is agent-ready. Does nothing on a device that is not parked.

Input parameters:

- `device` (string): Device id or name (optional when exactly one device is booted)

### `set_profile` (~91 tokens)

Switch the screen to a real-device profile (see profiles) or a custom Android size.

Input parameters:

- `density` (number): Custom density dpi (Android)
- `device` (string): Device id or name (optional when exactly one device is booted)
- `height` (number): Custom height px (Android)
- `profile` (string): Profile name
- `width` (number): Custom width px (Android)

### `screenshot` (~84 tokens)

Capture the screen. Returns the image plus its file path. Coordinates on a scale=1 image are tap coordinates.

Input parameters:

- `device` (string): Device id or name (optional when exactly one device is booted)
- `format` (string): png or jpeg
- `scale` (number): Downscale factor 0-1 for the returned image (default 0.5)

### `ui_tree` (~71 tokens)

Accessibility/UI hierarchy as a compact outline: [ref] role "text" id=... @(x,y wxh). Frames are screenshot pixels.

Input parameters:

- `device` (string): Device id or name (optional when exactly one device is booted)
- `json` (boolean): Return full JSON instead of the outline

### `find` (~120 tokens)

Find elements by text/label/id/role.

Input parameters:

- `contains` (string): Case-insensitive substring of text or label
- `device` (string): Device id or name (optional when exactly one device is booted)
- `id` (string): resource-id (Android) or accessibilityIdentifier (iOS)
- `index` (number): Pick the n-th match (0-based)
- `role` (string): Normalized role: button, text, textfield, switch, image, cell...
- `text` (string): Exact text or accessibility label

### `tap` (~259 tokens)

Tap at x,y (screenshot pixels) or at the center of the element matching a selector. With waitFor it also waits for the next screen.

Input parameters:

- `contains` (string): Case-insensitive substring of text or label
- `device` (string): Device id or name (optional when exactly one device is booted)
- `id` (string): resource-id (Android) or accessibilityIdentifier (iOS)
- `index` (number): Pick the n-th match (0-based)
- `returnUi` (boolean): Also return the UI outline of the screen the action ends on (act, wait and observe in one call)
- `role` (string): Normalized role: button, text, textfield, switch, image, cell...
- `text` (string): Exact text or accessibility label
- `waitFor` (object): After the action, wait for an element matching this selector (fields: text, contains, id, role) in the same call
- `waitGone` (boolean): Wait for waitFor to disappear instead
- `waitTimeoutMs` (number): Timeout for waitFor, default 10000
- `x` (number): X in screenshot pixels
- `y` (number): Y in screenshot pixels

### `swipe` (~191 tokens)

Swipe from (x,y) to (x2,y2) in screenshot pixels.

Input parameters:

- `device` (string): Device id or name (optional when exactly one device is booted)
- `durationMs` (number): duration, default 300
- `returnUi` (boolean): Also return the UI outline of the screen the action ends on (act, wait and observe in one call)
- `waitFor` (object): After the action, wait for an element matching this selector (fields: text, contains, id, role) in the same call
- `waitGone` (boolean): Wait for waitFor to disappear instead
- `waitTimeoutMs` (number): Timeout for waitFor, default 10000
- `x` (number, required): start x
- `x2` (number, required): end x
- `y` (number, required): start y
- `y2` (number, required): end y

### `type_text` (~175 tokens)

Type text into the focused field.

Input parameters:

- `device` (string): Device id or name (optional when exactly one device is booted)
- `keys` (boolean): Send one key event per character (slower) instead of setting the field's text (Android) or pasting it (iOS)
- `returnUi` (boolean): Also return the UI outline of the screen the action ends on (act, wait and observe in one call)
- `text` (string, required): Text to type
- `waitFor` (object): After the action, wait for an element matching this selector (fields: text, contains, id, role) in the same call
- `waitGone` (boolean): Wait for waitFor to disappear instead
- `waitTimeoutMs` (number): Timeout for waitFor, default 10000

### `press_key` (~158 tokens)

Press a key: home, back, enter, delete, tab, escape, lock, volumeup, volumedown.

Input parameters:

- `device` (string): Device id or name (optional when exactly one device is booted)
- `key` (string, required): Key name
- `returnUi` (boolean): Also return the UI outline of the screen the action ends on (act, wait and observe in one call)
- `waitFor` (object): After the action, wait for an element matching this selector (fields: text, contains, id, role) in the same call
- `waitGone` (boolean): Wait for waitFor to disappear instead
- `waitTimeoutMs` (number): Timeout for waitFor, default 10000

### `install` (~51 tokens)

Install an app (.apk for Android, .app for iOS Simulator).

Input parameters:

- `device` (string): Device id or name (optional when exactly one device is booted)
- `path` (string, required): Absolute path

### `launch` (~209 tokens)

Launch an app by package name / bundle id. iOS also takes process arguments (e.g. an Expo dev client's --initialUrl), environment and relaunch.

Input parameters:

- `app` (string, required): Package name or bundle id
- `args` (array): Process arguments (iOS)
- `device` (string): Device id or name (optional when exactly one device is booted)
- `env` (object): Process environment (iOS)
- `relaunch` (boolean): Terminate a running instance first (iOS)
- `returnUi` (boolean): Also return the UI outline of the screen the action ends on (act, wait and observe in one call)
- `waitFor` (object): After the action, wait for an element matching this selector (fields: text, contains, id, role) in the same call
- `waitGone` (boolean): Wait for waitFor to disappear instead
- `waitTimeoutMs` (number): Timeout for waitFor, default 10000

### `terminate` (~44 tokens)

Force-stop an app.

Input parameters:

- `app` (string, required): Package name or bundle id
- `device` (string): Device id or name (optional when exactly one device is booted)

### `wait_for` (~150 tokens)

Wait until an element matching the selector appears (or disappears with gone=true).

Input parameters:

- `contains` (string): Case-insensitive substring of text or label
- `device` (string): Device id or name (optional when exactly one device is booted)
- `gone` (boolean): Wait for disappearance
- `id` (string): resource-id (Android) or accessibilityIdentifier (iOS)
- `index` (number): Pick the n-th match (0-based)
- `role` (string): Normalized role: button, text, textfield, switch, image, cell...
- `text` (string): Exact text or accessibility label
- `timeoutMs` (number): Timeout, default 10000

### `open_url` (~198 tokens)

Open a URL or deep link as a link tapped in another app would (Android: VIEW intent with CATEGORY_BROWSABLE). App/universal links open the app only if their domain is verified. With waitFor it also waits for the next screen.

Input parameters:

- `device` (string): Device id or name (optional when exactly one device is booted)
- `returnUi` (boolean): Also return the UI outline of the screen the action ends on (act, wait and observe in one call)
- `url` (string, required): URL with a scheme: https://..., myapp://path, exp+app://...
- `waitFor` (object): After the action, wait for an element matching this selector (fields: text, contains, id, role) in the same call
- `waitGone` (boolean): Wait for waitFor to disappear instead
- `waitTimeoutMs` (number): Timeout for waitFor, default 10000

### `permission` (~188 tokens)

Grant, revoke or reset an app's privacy permissions without the system dialog. revoke = denied without asking again; reset = ask again on next use. Android kills the app when one is revoked or reset. A grant the app could not get on a device (not in its manifest / no Info.plist usage description) is refused.

Input parameters:

- `action` (string, required): What to do
- `app` (string, required): Package name or bundle id
- `device` (string): Device id or name (optional when exactly one device is booted)
- `permissions` (array): Names: camera, microphone, photos, photos-add, location, location-always, contacts, contacts-limited, calendar, reminders, notifications, motion, media-library, siri; or a raw Android permission (and…

### `location` (~87 tokens)

Set the simulated GPS location, or clear it (back to the device's baseline).

Input parameters:

- `action` (string): set (default when coordinates are given) or clear
- `device` (string): Device id or name (optional when exactly one device is booted)
- `latitude` (number): Degrees, -90..90
- `longitude` (number): Degrees, -180..180

### `clipboard` (~78 tokens)

Read the device clipboard (plain text), or set it.

Input parameters:

- `action` (string): get (default) or set (default when text is given)
- `device` (string): Device id or name (optional when exactly one device is booted)
- `text` (string): Text to put on the clipboard (action set; empty clears it)

### `appearance` (~53 tokens)

Read the system appearance, or switch light/dark mode.

Input parameters:

- `device` (string): Device id or name (optional when exactly one device is booted)
- `mode` (string): Omit to read the current appearance

### `apps` (~38 tokens)

List installed user apps (not the system image's).

Input parameters:

- `device` (string): Device id or name (optional when exactly one device is booted)

### `uninstall` (~48 tokens)

Uninstall a user app with its data.

Input parameters:

- `app` (string, required): Package name or bundle id
- `device` (string): Device id or name (optional when exactly one device is booted)

### `clear_app_data` (~77 tokens)

Clear one app's data so it starts like a fresh install (stopped, data and permissions gone; iOS keeps keychain items, as a reinstall does on a device).

Input parameters:

- `app` (string, required): Package name or bundle id
- `device` (string): Device id or name (optional when exactly one device is booted)

### `push` (~211 tokens)

Send a push notification to an app (iOS; Android has no faithful path). It shows only if the app may post notifications. With waitFor it also waits for the banner or screen.

Input parameters:

- `app` (string, required): Bundle id
- `device` (string): Device id or name (optional when exactly one device is booted)
- `payload` (object, required): APNs payload with an "aps" object, at most 4096 bytes, e.g. {"aps":{"alert":{"title":"Hi","body":"Hello"}}}
- `returnUi` (boolean): Also return the UI outline of the screen the action ends on (act, wait and observe in one call)
- `waitFor` (object): After the action, wait for an element matching this selector (fields: text, contains, id, role) in the same call
- `waitGone` (boolean): Wait for waitFor to disappear instead
- `waitTimeoutMs` (number): Timeout for waitFor, default 10000

### `logs` (~305 tokens)

Recent log lines of one app (or the whole device), oldest first, as `time level process[pid] tag: message`. Filter by time, minimum level and text; the newest lines that fit limit and maxBytes are returned. The last line gives a cursor: pass it as since to read only newer lines.

Input parameters:

- `app` (string): Package name or bundle id: only its processes' lines (default: every process)
- `contains` (string): Case-insensitive substring of the tag or the message
- `device` (string): Device id or name (optional when exactly one device is booted)
- `json` (boolean): Return JSON instead of text lines
- `level` (string): Minimum level (iOS: debug lines are not captured; info includes iOS default)
- `limit` (number): At most this many lines, the newest (default 100, max 10000)
- `maxBytes` (number): At most about this many bytes of lines (default 16384, max 1048576)
- `regex` (string): RE2 regular expression matched against the tag or the message
- `since` (string): Only lines from this point: a time (2026-11-07T10:00:00Z), a duration back from now (30s, 5m), "launch" (the app's last launch through agemu) or a cursor from a previous logs result (c123...)

### `crashes` (~172 tokens)

Crash, ANR (Android), system-kill (iOS watchdog, jetsam) and unexpected-exit reports of the device's apps, oldest first: time, kind, exception, reason and the crashing thread's top frames.

Input parameters:

- `app` (string): Package name or bundle id (default: every app)
- `device` (string): Device id or name (optional when exactly one device is booted)
- `json` (boolean): Return JSON instead of text
- `limit` (number): At most this many reports, the newest (default 10, max 100)
- `since` (string): A time (2026-11-07T10:00:00Z), a duration back from now (5m) or "launch" (the app's last launch through agemu)

## Diagnostics

Captured diagnostic sections: Provenance, Dependencies. The full working is on the page: https://verifymcp.io/servers/piotr-graczyk-dev-agemu/agentic-emulator#diagnostics

## Score history

- 2026-11-21: 82
- 2026-11-20: 82
- 2026-11-19: 81
- 2026-11-18: 81
- 2026-11-17: 80
- 2026-11-16: 80
- 2026-11-15: 64
- 2026-11-14: 64

## Common questions

### What is the agemu MCP server?

agemu is an MCP server listed in the public MCP registry as io.github.piotr-graczyk-dev/agemu. Fast Android emulator and iOS Simulator control for coding agents: screenshots, UI tree, taps, logs. This page covers its npm package (agentic-emulator).

### Is the agemu MCP server safe to use?

agemu scores 82 out of 100 on VerifyMCP. We found no known CVEs affecting it as of 21 November 2026. It declares no install or post-install scripts. Its build provenance is signed and verified. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

### What tools does the agemu MCP server expose?

agemu exposes 31 tools: devices, profiles, create, boot, shutdown, and 26 more. Their descriptions and schemas cost roughly 3,505 tokens of context every time the server is loaded.

### Is the agemu MCP server still maintained?

agemu is still listed as active in the MCP registry. We last reached this channel on 21 November 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

### What licence is the agemu MCP server under?

agemu declares the MIT licence, which is OSI-approved. That covers the source only, and says nothing about the cost of any service it calls.

### What did VerifyMCP test on the agemu MCP server?

VerifyMCP ran 16 checks on agemu, covering Malware scan, Known CVEs, Install scripts, Dependency health and 12 more. Endpoint Security and Transport & Reachability do not apply to an npm package, so they were not tested. The score does not measure everything, and the scoring documentation lists what it leaves out.

## Links

- npm package: https://www.npmjs.com/package/agentic-emulator
- Socket report: https://socket.dev/npm/package/agentic-emulator
- Repository: https://github.com/piotr-graczyk-dev/agentic-emulator
- Changelog RSS feed: https://verifymcp.io/servers/piotr-graczyk-dev-agemu/agentic-emulator.xml
- Changelog JSON feed: https://verifymcp.io/servers/piotr-graczyk-dev-agemu/agentic-emulator/changelog.json
- Trust report (JSON): https://verifymcp.io/servers/piotr-graczyk-dev-agemu/agentic-emulator.json
- HTML version of this page: https://verifymcp.io/servers/piotr-graczyk-dev-agemu/agentic-emulator
