Phonebox Android MCP server
OfficialProvides an agent-first cloud Android phone that an agent can control via tools: create and start/park phones, read the screen as a numbered text list with a screenshot, run ordered actions (tap, type, swipe, scroll, open apps, wait for elements), install apps from the app library or upload your own APK, reboot a stuck phone, and generate a live view link for human handoff. Phone state (apps, files, sign-ins) is preserved while parked.
This server has a remotely hosted endpoint. Open its connector listing for connection details, authentication requirements, and setup instructions:
Phonebox: cloud phones for AI agents (Android MCP server)
A hosted Android MCP server for Claude Code, Codex, Cursor and any MCP client, from Phonebox, the agent-first cloud phone. Your agent gets a real Android phone in the cloud. It can read the screen, tap, type, install your APK, hand off to a human, and park the phone when it's done. You don't need ADB, Android Studio or an emulator.
claude mcp add --transport http phonebox https://phonebox.dev/mcp \
--header "Authorization: Bearer $PHONEBOX_API_KEY"▶ Watch the 3-minute setup: Give Claude Code an Android Phone: Android MCP Server Setup · transcript ▶ Watch the 56-second demo: Give Claude Code an Android Phone (MCP) · transcript
Endpoint:
https://phonebox.dev/mcp(Streamable HTTP, bearer API key)Price: $0.06 per phone-minute, billed per second, 1-minute minimum. Parked phones are free and keep their apps and sign-ins.
Start free: $2 of credit, no card → phonebox.dev
This repo holds the configs and examples. The server itself is hosted by Phonebox, so there's nothing to run.
Why agent-first
Most cloud phones were built for social-media dashboards or QA test suites and added an API later. Phonebox was built for the agent: your agent signs you up from the terminal, reads the screen as text, gets errors with a next step, and never has a write retried behind its back. Idle phones park themselves for free.
Related MCP server: Karagöz
Why hosted instead of an ADB MCP server?
Local ADB MCP server | Phonebox hosted MCP | |
Where the phone runs | An emulator or USB phone on your machine | Phonebox's cloud |
Setup | Android SDK, ADB, a system image | One URL and an API key |
Works from CI, sandboxes and remote dev boxes | Only with nested virtualization or a device | Yes |
State between sessions | Whatever your emulator keeps | Apps, files and sign-ins kept while parked |
Human handoff | Someone at the machine | A live view link in any browser |
Get an API key
The fastest way is to paste this into your coding agent. It signs you up from the terminal, and you never paste a key into the chat:
Read https://phonebox.dev/setup.md and set me up with Phonebox from here: get me signed in and ready, then find out what I want to use phones for and help me with that.Or create a key in the console and export PHONEBOX_API_KEY=pbx_….
Agent skill
The Phonebox skill teaches a coding agent the whole workflow: onboarding, creating and parking phones, reading the screen, acting on it and installing apps. Install it with skills:
npx skills add combined-ai/phonebox-android-mcpOr point your agent at skills/phonebox/SKILL.md. The same file is served at https://phonebox.dev/skills/phonebox/SKILL.md.
Connect your client
Claude Code — run the command at the top, or commit this repo's .mcp.json. It reads the key from each person's environment:
{
"mcpServers": {
"phonebox": {
"type": "http",
"url": "https://phonebox.dev/mcp",
"headers": { "Authorization": "Bearer ${PHONEBOX_API_KEY}" }
}
}
}Cursor: add examples/cursor-mcp.json to ~/.cursor/mcp.json or .cursor/mcp.json. Guide: give Cursor an Android phone.
Codex: add examples/codex-config.toml to ~/.codex/config.toml and export PHONEBOX_API_KEY=$(phonebox token). Guide: give Codex an Android phone.
Gemini CLI:
gemini mcp add --transport http --scope user \
--header "Authorization: Bearer $PHONEBOX_API_KEY" \
phonebox https://phonebox.dev/mcpWindsurf, Cline and VS Code (GitHub Copilot): setup pages for Windsurf, Cline and GitHub Copilot.
Any other client: point it at https://phonebox.dev/mcp with an Authorization: Bearer pbx_… header. The server is stateless Streamable HTTP and doesn't use OAuth.
Tools
Tool | What it does |
| Creates a cloud Android phone and waits until it's ready |
| Reads the screen as a numbered text list, plus a 720px screenshot |
| Runs up to 20 actions in order: tap, type, swipe, scroll, open an app, wait for an element |
| Installs from the app library or from your own uploaded APK |
| Uploads your APK with one |
| A link for a person to watch and control the phone |
| Resumes a parked phone, or parks it to stop billing |
| Find phones, wait for them, list apps, restart a stuck phone |
What the agent reads from observe:
app com.android.launcher3
screen 1080x2400
[1] EditText "Search apps" #search clickable editable (540,250)
[2] TextView "Chrome" #icon clickable (200,1900)
[3] TextView "Settings" #icon clickable (500,1900)Examples
Test the APK you just built: build, install, explore, and report bugs with the exact screen text.
TypeScript SDK agent: create a phone, read the screen, tap, park.
A phone per customer: find or create each customer's phone by your own ID.
What people use it for
Testing the Android app the agent just built. Build the APK, install it, walk the changed screens, report bugs with the exact screen text. Guide · copy-paste test prompts · reproduce a bug report with a screen recording
Using apps that have no API. The agent reads the screen and taps through the app like a person, on its own phone, and hands you a live view link for payments and one-time codes. Guide
A phone for each customer of an AI product. One persistent phone per end user, tagged with your customer ID, parked between tasks so idle customers cost nothing. Guide · example
FAQ
How can I give Claude Code an Android phone?
Run the claude mcp add command at the top with a Phonebox API key. Claude Code gets a cloud Android phone it can read and control, with nothing to install. If you'd rather run everything locally, an emulator plus adb or a local server such as mobile-mcp also works; every option, compared.
Is there a hosted Android MCP server I can use with Claude Code or Cursor?
Yes, this one: https://phonebox.dev/mcp, Streamable HTTP with a bearer API key. The phones run in Phonebox's cloud, so it works from CI runners and cloud sandboxes where no emulator can run.
How is it different from mobile-mcp or Appium MCP? Those run on your machine and drive an emulator or USB device you provide, and they also cover iOS. Phonebox brings the phone, keeps it signed in between sessions, and is Android only. Comparison.
How can an AI agent use Android apps that have no API? By using the app's screen: the agent reads each screen as a numbered list of elements, taps and types by label, and reads again after every action. Phonebox gives it a phone to do that on, and a live view link when a person must step in.
What does it cost? $0.06 per phone-minute while a phone runs, billed per second with a 1-minute minimum. Parked phones are free. A 10-minute test pass is $0.60. New accounts get $2 of credit, no card.
What it isn't
Android only (no iOS), and there are no phone numbers, SMS or calls. Phonebox gives your agent the device, not telephony.
Learn more
Phonebox (phonebox.dev) is a cloud phone platform for AI agents. It is not a mobile carrier, phone plan or eSIM provider.
Related MCP Connectors
Cloud Android phones for AI agents: read the screen, tap, type, install APKs, park and resume.
Give your AI agent a real cloud Android phone: push APKs, drive it, get screenshots and logs.
Let AI agents control Android and iOS devices on Mobilerun Cloud: screen, input, apps, workflows.
Disposable cloud Android emulators for coding agents: run an APK or PR build, tap, type, screenshot.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceEnables AI assistants to remotely control Android devices in the cloud via P2P WebRTC, providing tools for screenshots, touch input, file management, and automation scripts.13 npmMIT
- AlicenseBqualityBmaintenanceEnables AI agents to automate Android and iOS devices, including emulators and physical devices, by capturing screenshots, reading UI accessibility trees, performing touch and text input, and managing app lifecycle and logs.184MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to safely use real Android phones by listing devices, reading screens and UI elements, tapping, typing, launching apps, and running workflows only within allowed permissions and approvals, with every step recorded in an audit log.40 npmMIT
- AlicenseAqualityBmaintenanceA cloud Android phone for AI agents: push an APK and drive a real device — tap, type, screenshots, UI hierarchy, logs — while the human watches the same phone through a preview link.27MIT