android-mcp-toolkit
A growing collection of MCP tools for Android Development. Currently features a deterministic Figma-SVG-to-Android-XML converter, with plans for Gradle analysis, Resource management, and ADB integration tools.
Android MCP Toolkit for AI Agents
Small MCP server with three tools:
- Fast SVG → Android VectorDrawable conversion (cached, file or inline).
- adb logcat reader with package/pid/tag filters for quick crash triage.
- Translation length difference estimator to flag risky length deltas before layout breaks.
Why this exists
The Mission: Bringing Native Android to the AI Agent Era
While the AI ecosystem flourishes with web-first tools, Android development often feels left behind. This MCP server is my answer to that gap—a dedicated bridge connecting AI Agents directly to the Android toolchain.
⚡ Zero-Friction Asset Conversion: Convert SVGs to VectorDrawables instantly without the overhead of launching Android Studio.
🔍 Direct Diagnostic Access: Empower agents to pull, filter, and analyze adb logcat streams (by package, PID, or tag) in real-time.
🤖 Agent-Native Architecture: Deliver structured, scriptable outputs that LLMs can parse and act upon efficiently.
🚀 Built for Extensibility: A solid foundation designed to grow, paving the way for future utilities like bitmap helpers and asset validation.
Pairing ideas
- Figma MCP: grab SVGs from designs, feed to
convert-svg-to-android-drawableto get XML for Android resources. - Debugging: while running the app, call
read-adb-logcatwith package name or tag to capture crashes and filtered logs without leaving the MCP client.
Previews
SVG to VectorDrawable
- Figma request → SVG extraction

- Flag conversion preview (single)

- Batch flag review (side-by-side)

- Batch run via MCP (console)

ADB logcat tool
- Crash capture prompt (inputs + filters)

- Response preview (summarized logcat)

Current tools
-
convert-svg-to-android-drawable- Inputs:
svg(inline) orsvgPath(file path). Optional:outputPath,floatPrecision(default 2),fillBlack(default false),xmlTag(default false),tint,cache(default true). - Output: VectorDrawable XML text; also writes to disk when
outputPathis provided. - Performance: LRU cache (32 entries) keyed by SVG + options plus fast reuse in-session.
- Converter: vendored fork in
vendor/svg2vectordrawablewith fixes forrgb()/rgba(),hsl()/hsla(), and named colors. Upstream license:vendor/svg2vectordrawable/LICENSE(MIT).
- Inputs:
-
manage-logcat- Inputs:
action:read(default),crash,anr, orclear.packageName: Optional. Resolves PID viaadb shell pidof.pid: Optional. Explicit PID.tag: Optional. Filter by tag (e.g.MyApp).priority:V,D,I,W,E,F,S(defaultV).maxLines: Tail count (default 200, max 2000).timeoutMs: Default 5000.
- Behavior:
read: Fetches logcat tail.crash: Fetcheslogcat -b crash.anr: Fetches recent ActivityManager ANR logs + tail of/data/anr/traces.txt.clear: clears logcat buffers.
- Inputs:
-
get-current-activity- Inputs:
timeoutMs(default5000, max15000). - Behavior: Inspects
dumpsys windowto find the currently focused app/window. Useful to verify state.
- Inputs:
-
dump-ui-hierarchy- Inputs:
timeoutMs(default 10000). - Behavior: Captures current UI hierarchy as XML via
uiautomator.
- Inputs:
-
take-screenshot- Inputs:
outputPath(required),timeoutMs. - Behavior: Saves device screenshot to local file.
- Inputs:
-
inject-input- Inputs:
command(tap,text,swipe,keyevent,back,home),args(array),timeoutMs. - Optional:
elementIdorelementText(finds element center and taps it). - Behavior: Simulates user interaction suitable for testing flows.
- Inputs:
-
estimate-text-length-difference- Inputs:
sourceText(original),translatedText(to compare),tolerancePercent(default30, max500). - Behavior: Measures grapheme length of both strings, computes percent change, and reports whether it exceeds the tolerance (useful to catch translation length blowups that could break layouts).
- Inputs:
Roadmap (planned)
- Additional MCP tools for Android assets (e.g., batch conversions, validations, optimizers).
- Optional resource prompts for common Android drawables/templates.
Quick start
npm installnpm run buildnode dist/index.js(stdio MCP server)
Run via npx
- Global:
npx android-mcp-toolkit
Use in Cursor (MCP config)
Add to your Cursor settings JSON:
{
"mcpServers": {
"android-mcp-toolkit": {
"command": "npx",
"args": [
"-y",
"android-mcp-toolkit"
]
}
}
}
The npx call downloads the published package; no local path required.
Quick install via Cursor deep link:
cursor://anysphere.cursor-deeplink/mcp/install?name=android-mcp-toolkit&config=eyJjb21tYW5kIjoibnB4IC15IGFuZHJvaWQtbWNwLXRvb2xraXQifQ%3D%3D
Examples
- Input SVG:
sample_svg.svg - Output VectorDrawable:
examples/sample_svg.xml
Notes
- Transport: stdio via
@modelcontextprotocol/sdk. - Base deps kept minimal; everything needed to convert SVGs is vendored/included.
Contact
เซิร์ฟเวอร์ที่เกี่ยวข้อง
Scout Monitoring MCP
ผู้สนับสนุนPut performance and error data directly in the hands of your AI assistant.
Alpha Vantage MCP Server
ผู้สนับสนุนAccess financial market data: realtime & historical stock, ETF, options, forex, crypto, commodities, fundamentals, technical indicators, & more
onUI
Annotate elements, draw regions, and ship cleaner UI faster.
Whistle MCP Server
Manage local Whistle proxy servers through AI assistants, simplifying network debugging, API testing, and rule management.
AppleScript MCP
Execute AppleScript on macOS
FAL Imagen 4
Generate high-quality images using Google's Imagen 4 Ultra model via the FAL AI platform.
Luma AI MCP Server
Generate and manipulate AI videos and images using Luma AI's Dream Machine API.
Godot MCP
MCP server for interacting with the Godot game engine, providing tools for editing, running, debugging, and managing scenes in Godot projects.
FeedOracle Macro Intelligence
Deterministic macro regime engine for AI trading agents — RISK_ON / NEUTRAL / RISK_OFF / STRESS with transparent scoring rules.
MCP Server for Drupal
A TypeScript-based MCP server that acts as a companion to the Drupal MCP module, communicating via STDIO.
AgentExecMCP
A secure, Docker-based server providing core execution capabilities for AI agents.
Agent Loop
An AI Agent with optional Human-in-the-Loop Safety and Model Context Protocol (MCP) integration.