Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Description from the maker (original English)
Let an AI agent see and use a real Android phone over Wi-Fi, without ADB or root.
These three are never merged into one score: opinions, measurements and popularity are different things.
We have not written a plain-language summary for this one yet. Below are the facts we read from the registry.
Pick the AI app you use, then copy the settings.
Which app do you use?
Pick your app above. We remember it, so next time the button says “Install in …”.
Download the maker’s .mcpb file
The link goes to the maker’s own file; we do not host or repackage it. Open the file and Claude Desktop installs it.
Run this in your terminal
Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
This one runs on your computer, and ChatGPT can only connect to servers on the internet. Pick another app.
Your choice stays on this device in a cookie; no account needed.
These settings are generated from the maker’s registry entry. If the README says otherwise, follow the README.
Runs on your computer; the tool list appears in your AI app after you install it.
We check information anyone can see; we have not read the code. Checks that do not apply are left out of the count. · Checked Nov 6, 2026
The registry name was verified with the GitHub account KnifeLemon, which owns the repo
We have not read this repo from GitHub yet
We have not read this repo from GitHub yet
This one runs on your computer; there is no internet endpoint
Runs on your computer and uses local settings instead
Runs on your computer; we can only read tool lists from internet endpoints
Compared with every active server in this category in the registry. Stars count only when the publisher owns the repo.
It passed 1 of 1 applicable basic checks. We have not read the code. Before adding a key or token, see what access the server asks for and use a key with only the permissions you need.
The registry does not list a required key, though some servers ask you to sign in on first use.
It runs on your computer. Your AI app starts it when needed, and data does not pass through the maker’s servers (unless it calls an online service itself).
The GitHub repo last changed –, and the latest version (0.1.13) reached the registry 3 days ago.
Apps that support MCP, such as Claude Desktop, Claude Code, Cursor and VS Code. You also need MCPB on your computer.
In the same category: Spec Driven Development, Remote Desktop Commander, Reelsaw
Servers in the same category from other makers, sorted by stars.
Reading a card:★ GitHub starsiThree pillars: reviews · checks · usageiUpdated = latest version published to the registryi
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Drive a self-hosted Reelsaw: turn long videos into vertical clips, edit, re-render and export them.
Copied from the maker’s repo as written. PopNx MCP has not reviewed or endorsed it. HTML, images and scripts are removed. Links lead off this site.
Automate real Android phones from your code.
Test your apps, script the tasks you repeat, run a shelf of phones, or let an AI agent use one.
Python, Node.js, the shell or HTTP. No ADB, no USB cable, no root.
Website · Docs · Download · Sponsor · Quick start · Languages · Guides
English · 한국어 · 简体中文
No ADB, no cable Install one app on the phone, turn on its permissions and pair it with a QR code. No developer options, no USB debugging, no root. The same command everywhere touch is touch in Python, Node.js, the CLI, HTTP and as an MCP tool, all generated from one spec. Phones anywhere The phone connects out to your PC, so it works on the same Wi-Fi or on mobile data through a tunnel or a relay you run. Every line is encrypted end to end.
Droidline uses only permissions an ordinary app can get, so a few things are out of reach: see What it cannot do.
Try the same demo in your browser, and pick elements on a simulated screen, on droidline.dev
You need a PC (Windows, macOS or Linux) and an Android phone with Android 9 or later, on the same Wi-Fi for a first try. The installation guide walks through each step with screenshots.
droidline archive for your system from Releases, put droidline on your PATH, and start it. Leave it running. droidline serve
droidline-agent.apk from the same release page and open it. On the app's Setup tab, turn on the accessibility service, the Droidline keyboard, notifications and unrestricted battery use. On Android 13 and later, allow restricted settings for the app first; the app shows where. droidline pair
No camera? Tap Pair on this Wi-Fi in the app and type the 6-digit code it shows: droidline pair 482913.
droidline launch com.android.settings
droidline touch text "Network & internet"
No phone at hand? droidline-fakephone, included in every release, simulates one with a small demo app. The first-script tutorial builds a complete login script against it.
Every interface talks to droidline serve on localhost:8780 and uses the same command names.
Python (pip install droidline, Python 3.9+)
from droidline import connect
d = connect() # the phone paired with this PC d.launch("dev.droidline.demo") if d.exists("text", "Close ad"): # conditions never wait and never fail d.touch("text", "Close ad") d.input("id", "email", "knife") # waits up to 10 s for the element d.touch("text", "Log in") print(d.get_text("id", "greeting"))
Node.js and TypeScript (npm install droidline, Node.js 18+)
import { connect } from "droidline";
const d = await connect(); await d.launch("dev.droidline.demo"); await d.input("id", "email", "knife"); await d.touch("text", "Log in", { timeout: 15 }); console.log(await d.getText("id", "greeting"));
C# and .NET (dotnet add package Droidline, .NET 8+ or .NET Framework 4.6.2+)
using Droidline;
await using var d = await DroidlineClient.ConnectAsync(); await d.LaunchAsync("dev.droidline.demo"); await d.InputAsync("id", "email", "knife"); await d.TouchAsync("text", "Log in", timeout: 15); Console.WriteLine(await d.GetTextAsync("id", "greeting"));
Command line (comes with the server)
droidline touch text "Log in"
droidline which text="Log in" id=main_tab --timeout 15 droidline screenshot shot.png
HTTP (any language, any tool)
curl -s -X POST localhost:8780/devices/_/touch \
-H 'content-type: application/json' -d '{"by":"text","value":"Log in"}'
AI agents over MCP (Claude, Cursor, VS Code and others)
{ "mcpServers": { "droidline": { "command": "droidline", "args": ["mcp"] } } }
With Claude Code: claude mcp add droidline -- droidline mcp. In Claude Desktop you can instead open droidline.mcpb from Releases, which carries its own droidline; it is also listed in the MCP Registry as io.github.KnifeLemon/droidline. To give an agent fewer tools, add --tools dump,screenshot,touch or --read-only to the args; see Limit what an agent can use. Other languages can open a TCP socket and send one JSON line per command; see Other languages for Go, Java and PHP.
dump() returns the screen as a tree with text, id and desc for every element. Commands take the field and the value: touch("id", "login"). When an element refuses a click, Droidline taps its parent or the center of its bounds by itself.touch, input and wait wait for their element, 10 seconds by default. Conditions such as exists, checked and which answer at once and never throw for a missing element.NOT_FOUND: Could not find text 'Log in' within 10s. Current screen: com.example / .MainActivity.batch sent with cuts_network keeps running on the phone while it is offline, so airplane mode on, wait, off works as one call. intent opens the Settings page first; the recipes show it step by step.touch({"class": "android.widget.Switch", "row": {"text": "Wi-Fi"}}). find hands back elements you can click and search inside, and wait_idle waits until the screen stops changing.droidline inspect shows the screen and its elements, suggests selectors with code, and records a script while you use the phone. droidline webdriver lets Appium clients and scripts drive your phones. There is also a pytest plugin, leases for sharing phones between scripts, and picture and text matching for screens without elements.Setup: each permission with a button to its settings screen Pair: scan the QR code or use a 6-digit code Status: connection, route and permissions
The phone always opens the connection, so it needs no open port. Pick the route that fits your network; the phone tries them in order and returns to Wi-Fi when it can.
| Route | You need | How it works |
|---|---|---|
| Same Wi-Fi | Nothing | The phone finds the PC with a UDP broadcast and connects directly. |
| Port forward | One forwarded port | The phone connects to your public address over TLS, pinned to the PC's certificate. |
| Tunnel | cloudflared, ngrok or similar | Your PC keeps a tunnel open; the phone reaches it through the tunnel's address. |
| Your relay | Cloudflare Workers or a VPS | PC and phone both connect out to a relay you deploy. It forwards lines it cannot read. |
Past the two-line handshake every line is encrypted with keys only your PC and phone hold (P-256 and AES-256-GCM), so a tunnel or relay only carries ciphertext. The remote connection guides cover each route step by step.
Droidline only uses permissions an ordinary app can get, so some things are out of reach. Better to know before you start:
launch opens the start screen instead.Only the beginning is shown; read the rest in the repo