Claude Code
Lets you hand coding tasks to Anthropic's AI agent in your terminal, editor or desktop app
by Anthropic
agent-device
by CallstackCLI toolWrite code
Lets your coding agent tap through a mobile or desktop app and save screenshots as proof
About agent-device
agent-device is a command-line tool and MCP server that lets AI coding agents inspect, control and check mobile, TV, desktop and web apps, and save screenshots, video and logs as evidence.
For Mobile app developersQA engineersCoding-agent users
Made by Callstack
agent-device is made at Callstack, which publishes it as open source under the MIT licence and offers team QA help on request.
- Stars
- ★ 5.3k
- License
- MIT
- Cost
- Free
- Platforms
- Mac · Win · Linux
- Level
- Advanced
Overview
agent-device is a command-line tool from Callstack that lets a coding agent open, inspect, control and check apps on iOS, Android, TV, macOS, Linux and the web, on simulators, emulators and real devices. The agent reads the app's accessibility tree, taps and types through it, and saves screenshots, video and logs as evidence. It also runs as an MCP server. It suits app developers who already use a coding agent.
The agent can check its own app changes on a running simulator and attach screenshots and logs to a pull request.
Try these first
Start with the first one: it is a small task you can check easily.
- Open the Settings app on the iOS simulator, take a screenshot and close the session.First try · read-only
- Run [my app] on the Android emulator and attach screenshots of the onboarding screen.saves files
- Reproduce this crash in [my app] and capture the logs that lead up to it.
- Explore the checkout flow once, save it as a replay script and run it again.writes a script
Example prompts written by Agent Flows Hub.
Features
From the creator's documentation.
| Feature | What it does | Runs |
|---|---|---|
| Inspect the screen | Reads the app's accessibility tree as compact snapshots with element references | Your computer |
| Act on the app | Taps, fills fields, scrolls, makes gestures, waits and handles alerts | Your devices |
| Evidence | Saves screenshots, video, logs, traces, network data and crash details | Your computer |
| Replay scripts | Saves working steps as .ad scripts for local use or CI, or exports Maestro YAML | Your computer |
| MCP server | Starts a stdio MCP server that exposes the same commands as tools | Your computer |
| Device clouds | Connects to BrowserStack, AWS Device Farm, TestMu AI and Limrun | Cloud services |
Good to know
You build a mobile app and use a coding agent. After the agent changes the app, agent-device lets it run the app on a simulator, tap through a screen and save screenshots and logs you can review.
Limits
- It needs the platform tools of each target on your machine, such as Xcode for iOS and the Android SDK with ADB for Android; web automation needs Node.js 24 or newer.
- Support depth varies by target; the capabilities command shows what each platform supports.
Notes by Agent Flows Hub (original project name: agent-device), written from the creator's documentation, not from running it ourselves. agent-device installation guide · agent-device README
Stuck setting it up?
Tell us what happened. We take a look first; if it needs paid work, you get an estimate before anything starts.
More to write code
Other cli tools
- ai-memory Gives your coding agents a shared memory so the next one picks up where you stopped
- ccusage Shows how many tokens your coding agents used and what that would cost, day by day
- Claude Code Templates Installs ready-made skills, agents, commands and MCP setups into Claude Code in one step
- Claude Code Usage Monitor Shows your Claude Code usage live in the terminal and warns you before you hit a limit
- Claude Token Optimizer Shrinks the notes Claude Code rereads at every session start, so each uses fewer tokens
- CodeBurn Shows what your AI coding tools cost, split by project, model and task