Agent Device
ActiveAgent Device is Callstack's open-source command-line tool and MCP server that lets AI coding agents inspect, control, and verify mobile, TV, desktop, and web applications.
Overview and fit
Audience
Editorial judgment, not a vendor fact.
Compatibility
- Test languages
- JavaScript TypeScript Natural language
- Device modes
- Real device Emulator Simulator
What it tests
- Product kind
- Framework, AI testing agent
- Testing types
- End-to-end, Functional, Regression, Exploratory
- Targets
- Web app, Native desktop, Native mobile
- Use cases
- Regression testing, Pull request validation, Test generation
Classification lists positively established capabilities. Absence from a list does not mean lack of support.
Creation and execution
- Authoring methods
- Code, AI-generated workflow, Autonomous exploration, Imported tests
- Execution models
- OS input, Real device, Emulator, Simulator, Hybrid
- Deployment
- Local, Self-hosted
- License
- Open source (MIT)
Capabilities
AI
| Capability | Status | Details |
|---|---|---|
| Uses AI | Supported Supported | |
| AI test generation | Supported Supported | |
| AI executes the app | Supported Supported | |
| AI generates code only | Unsupported Unsupported | Null |
| Autonomous exploration | Supported Supported | |
| Vision-based interaction | Supported Supported | |
| Selector-free execution | Supported Supported | |
| Adapts to UI changes | Unknown Unknown | |
| Self-healing | Unknown Unknown | |
| AI failure analysis | Supported Supported |
Developer integrations
| Capability | Status | Details |
|---|---|---|
| MCP server | Supported Supported | |
| VS Code extension | Unknown Unknown | |
| SDK | Supported Supported |
PR workflow
| Capability | Status | Details |
|---|---|---|
| GitHub | Supported Supported | |
| CI execution | Supported Supported | |
| GitHub PR checks | Unknown Unknown | |
| PR-aware test selection | Unknown Unknown | |
| Autonomous PR validation | Supported Supported | |
| PR comments | Unknown Unknown | |
| Preview deployment testing | Unknown Unknown | |
| Merge gating | Unknown Unknown | |
| Scheduled execution | Unknown Unknown |
Creation and execution
| Capability | Status | Details |
|---|---|---|
| Code authoring | Supported Supported | |
| Recorder | Supported Supported | |
| Natural language authoring | Supported Supported | |
| Test import | Supported Supported | |
| Reusable steps | Supported Supported | |
| Parameterized tests | Supported Supported | |
| Test data generation | Unknown Unknown |
What it tests
| Capability | Status | Details |
|---|---|---|
| Native desktop testing | Supported Supported | |
| Native mobile testing | Supported Supported | |
| Browser extension testing | Unknown Unknown | |
| Cross-browser testing | Unknown Unknown | |
| Cross-platform testing | Supported Supported | |
| Real device testing | Supported Supported | |
| Multi-tab testing | Unknown Unknown | |
| iframe testing | Unknown Unknown | |
| Canvas testing | Unknown Unknown | |
| PDF validation | Unknown Unknown | |
| File upload testing | Unknown Unknown | |
| File download testing | Unknown Unknown | |
| OAuth testing | Unknown Unknown | |
| Email testing | Unknown Unknown | |
| Notification testing | Supported Supported | |
| Video content validation | Unknown Unknown |
Validation
| Capability | Status | Details |
|---|---|---|
| Visual regression | Supported Supported | |
| Semantic visual assertions | Supported Supported | |
| Accessibility checks | Unknown Unknown | |
| API assertions | Unknown Unknown | |
| Database assertions | Unknown Unknown | |
| Spelling and grammar checks | Unknown Unknown | |
| Dark and light mode testing | Unknown Unknown | |
| Performance measurement | Supported Supported |
Running
| Capability | Status | Details |
|---|---|---|
| Local execution | Supported Supported | |
| Headless execution | Unknown Unknown | |
| Headed execution | Supported Supported | |
| Parallel execution | Supported Supported | |
| Test sharding | Unknown Unknown | |
| Auto-waiting | Supported Supported | |
| Automatic retries | Supported Supported | |
| Network mocking | Unknown Unknown | |
| Secrets injection | Unknown Unknown | |
| Private network access | Unknown Unknown |
Diagnosis
| Capability | Status | Details |
|---|---|---|
| Screenshots | Supported Supported | |
| Video recording | Unknown Unknown | |
| Execution traces | Supported Supported | |
| Console logs | Supported Supported | |
| Network logs | Supported Supported | |
| Replay | Supported Supported | |
| Interactive debugging | Supported Supported | |
| Failure grouping | Unknown Unknown |
Enterprise
| Capability | Status | Details |
|---|---|---|
| SSO (SAML) | Unknown Unknown | |
| SSO (OIDC) | Unknown Unknown | |
| SCIM | Unknown Unknown | |
| RBAC | Unknown Unknown | |
| Audit logs | Unknown Unknown | |
| Secrets management | Unknown Unknown | |
| Configurable retention | Unknown Unknown | |
| Private runners | Unknown Unknown | |
| On-premises deployment | Supported Supported | |
| Data residency | Unknown Unknown | |
| Training opt-out | Unknown Unknown |
Integrations
| Capability | Status | Details |
|---|---|---|
| BrowserStack (Device cloud) | Supported Supported | Integration No source recorded |
| AWS Device Farm (Device cloud) | Supported Supported | Integration No source recorded |
| Limrun (Device cloud) | Supported Supported | Integration No source recorded |
Maintenance, reliability, and debugging
- Setup difficulty
- Low
- Maintenance approaches
- Manual code, Resilient locators
- Maximum concurrency
- null
- Debug artifacts
- Screenshot, Trace, Console log, Network log
- Reliability mechanisms
- Auto wait, Retry
- Interfaces
- CLI, API, SDK, MCP
Pricing
Free
Not yet verified
Security and deployment
- Certifications
- None recorded
- Data regions
- Unknown
- Data retention
- Unknown
- Customer data used for training
- Unknown
Support
- Channels
- Docs, Github issues
- Support SLA
- Unknown
Limitations
- iOS 'Allow Paste' dialog suppressed under XCUITest — Documented limitation on iOS
- Android non-ASCII text on real devices without the test IME helper — Documented limitation on Android
- Android first helper install can wait on an OEM install dialog — Documented limitation on Android
- Android no clipboard access over adb on Android 16 — Documented limitation on Android
SDK code samples
- Package:
- agent-device (version not recorded)
- Install:
- npm install agent-device
import { createAgentDeviceClient } from 'agent-device';
async function main() {
const client = createAgentDeviceClient({
session: 'qa-integration',
platform: 'ios',
udid: 'auto',
});
try {
await client.sessions.open();
await client.command.openApp({ bundleId: 'com.apple.Preferences' });
await client.command.tapOn({ text: 'Wi-Fi' });
await client.command.tapOn({ text: 'Other...' });
await client.command.inputText({ text: 'MyNetwork', into: 'Network Name' });
await client.command.tapOn({ text: 'Join' });
} finally {
await client.sessions.close();
}
}
await main();Run: node --experimental-strip-types examples/sdk/client-session.ts
Adapted from official docs
Sources and verification
- 1The mobile verification for AI Agents · Official website · Callstack · retrieved 2026-09-26
Supports 43 claims · Vendor documented, high confidence · verified 2026-09-26
- Accessibility checks
- Adapts to UI changes
- API assertions
- Browser engines
- Browser extension testing
- Browsers
- Canvas testing
- Cross-browser testing
- Cross-platform testing
- Dark and light mode testing
- Database assertions
- Deployment
- Email testing
- Failure grouping
- File download testing
- File upload testing
- Free tier
- Headed execution
- Headless execution
- iframe testing
- Local execution
- Multi-tab testing
- Native desktop testing
- Native mobile testing
- Network mocking
- OAuth testing
- On-premises deployment
- PDF validation
- Pricing
- Pricing models
- Private network access
- Product status
- Secrets injection
- Self-healing
- Spelling and grammar checks
- Support SLA
- Targets
- Test data generation
- Test sharding
- Tool
- Trial
- Vendor
- Video content validation
The mobile verificationfor AI Agents: Device automation CLI for AI agents. Real apps on iOS, Android, TV, and desktop.
Built for agents. Anywhere.: We believe all agents should have access to agent-device running real simulators and devices regardless whether they operate within your desktop or from a cloud sandbox.
The mobile verificationfor AI Agents: $npm install -g agent-device
The mobile verificationfor AI Agents: Device automation CLI for AI agents.
The mobile verificationfor AI Agents: Real apps on iOS, Android, TV, and desktop.
The mobile verificationfor AI Agents: The mobile verification for AI Agents. Device automation CLI for AI agents.
Clean-State Testing: agent-deviceMade by Callstack - React and React Native development experts
- 2GitHub - callstack/agent-device: Mobile app automation and verification for AI coding agents. CLI, MCP server, and typed Node.js API for iOS, Android, HarmonyOS, TV, web, macOS, and Linux. · GitHub · Official repository · Callstack · retrieved 2026-09-26
Supports 24 claims · Vendor documented, high confidence · verified 2026-09-26
- Audience
- Autonomous PR validation
- CI execution
- Company size
- Device modes
- Execution model
- GitHub PR checks
- License
- License identifiers
- Max concurrency
- Merge gating
- Parallel execution
- PR comments
- PR-aware test selection
- Preview deployment testing
- Product kind
- Real device testing
- Runner OS
- Scheduled execution
- Summary
- Support channels
- Target OS
- Testing types
- Use cases
Folders and files: Let your coding agent verify its changes in the running app.
Folders and files: agent-device is a command-line tool and MCP server that lets AI coding agents inspect, control, and verify mobile apps.
Folders and files: simulators, emulators, and physical devices
Folders and files: MIT license
Folders and files: It also coordinates device access across parallel agent worktrees
Folders and files: macOS, and Linux.
Folders and files: Docs
Folders and files: iOS, Android, and HarmonyOS (simulators, emulators, and physical devices), TV (tvOS, Android TV, Vega OS Virtual Device), web, macOS, and Linux.
Folders and files: Automated pull request and merge validation with replay scripts and captured artifacts.
- 3Node.js API - agent-device · Official docs · Callstack · retrieved 2026-09-26
Supports 6 claims · Vendor documented, high confidence · verified 2026-09-26
- Code authoring
- Code samples
- Integrations
- Interfaces
- Languages
- SDK
#Node.js API: Use createAgentDeviceClient() for typed, deterministic device automation from Node.js instead of shelling out to the CLI.
#Device cloud sessions: See the BrowserStack, AWS Device Farm, and Limrun guides for provider-specific setup.
- 4Replay & E2E Testing - agent-device · Official docs · Callstack · retrieved 2026-09-26
Supports 11 claims · Vendor documented, high confidence · verified 2026-09-26
- Authoring method
- Auto-waiting
- Automatic retries
- Maintenance approach
- Parameterized tests
- Recorder
- Reliability mechanisms
- Replay
- Reusable steps
- Semantic visual assertions
- Test import
#Run Maestro compatibility flows: Agent Device can run a supported subset of Maestro YAML through its typed Maestro runner.
#Run Maestro compatibility flows: extendedWaitUntil
- 5AI Agent Setup - agent-device · Official docs · Callstack · retrieved 2026-09-26
Supports 14 claims · Vendor documented, high confidence · verified 2026-09-26
- AI executes the app
- AI failure analysis
- AI generates code only
- AI test generation
- Autonomous exploration
- Console logs
- Debug artifacts
- Execution traces
- Interactive debugging
- MCP server
- Natural language authoring
- Network logs
- Uses AI
- VS Code extension
#Recommended agent rule: For logs, network, audio, traces, or runtime failures, read `agent-device help debugging`.
- 6Snapshots - agent-device · Official docs · Callstack · retrieved 2026-09-26
Supports 1 claim · verified 2026-09-26
- Vision-based interaction
- 7Selectors - agent-device · Official docs · Callstack · retrieved 2026-09-26
Supports 1 claim · verified 2026-09-26
- Selector-free execution
- 8Commands - agent-device · Official docs · Callstack · retrieved 2026-09-26
Supports 6 claims · verified 2026-09-26
- GitHub
- Notification testing
- Performance measurement
- Screenshots
- Video recording
- Visual regression
- 9Installation - agent-device · Official docs · Callstack · retrieved 2026-09-26
Supports 1 claim · Editorial inference, high confidence · verified 2026-09-26
- Setup difficulty
#Installation: #Global install
- 10Security & Trust - agent-device · Official docs · Callstack · retrieved 2026-09-26
Supports 13 claims · Vendor documented, high confidence · verified 2026-09-26
- Audit logs
- Certifications
- Configurable retention
- Customer data used for training
- Data regions
- Data residency
- Private runners
- RBAC
- SCIM
- Secrets management
- SSO (OIDC)
- SSO (SAML)
- Training opt-out
#Security & Trust: #Security & Trust
#Local control: #Local control
- 11Known Limitations - agent-device · Official docs · Callstack · retrieved 2026-09-26
Supports 1 claim · Vendor documented, high confidence · verified 2026-09-26
- Editorial · Limitations
#Known Limitations: #iOS: "Allow Paste" dialog suppressed under XCUITest
Hands-on tested: no (documentation research only) · Content updated 2026-09-26