Argent
NPM · @SWMANSION/ARGENT
Drive iOS, Android, HarmonyOS, TV, Mac and Electron/web apps from your coding agent
Available components
Last scanned (15 hours ago)
Scores higher than 79% of npm servers
Category Browser Automation
How this component scores in each security and reliability category. Every signal is checked automatically from public evidence about the published package, including repeated runs of it in an isolated sandbox, and we only credit what we can confirm. How we score → Why this is hard to score →
Supply Chain Security98
- No malware found by supply-chain analysis.Pass
- No known CVEs affecting this package version or its production dependencies.Pass
- No install/post-install scripts declared.Pass
- 64 of 177 dependencies flagged as unhealthy (1 deprecated). View diagnostics → Partial
Provenance & Transparency97
- Source repository is publicly reachable at the declared URL. View diagnostics → Pass
- Cryptographically verified build provenance (signed, bound to software-mansion/argent). View diagnostics → Pass
- Clear OSI-approved license (Apache-2.0).Pass
- Actively maintained (last published 1 days ago).Pass
- Disclosure check failed: no security disclosure policy was found in the source repository. See how to fix → Fail
Schema Quality & AI Usability61
- AI-judged instruction clarity (excellent).Pass
- Context-footprint check failed: tool/resource definitions use about 26173 tokens (~384/item across 68 items; 68 tools + 0 resources), over budget; trim descriptions and params. See how to fix → Fail
- Usage-examples check failed: none of the tools include examples. See how to fix → Fail
Stability & Change Management68
- Stability check failed: the tool surface changed between 0.19.0 and 0.34.0: 14 tool removals, 19 breaking changes, 9 additions. See how to fix → Fail
Tool Coverage100
- 100% of tools have a non-trivial description (not blank, and not just the tool's name).Pass
- 100% of tool parameters carry a description.Pass
Tool Safety75
- No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.Pass
- 0 of 11 tool(s) whose name or description implies an irreversible operation declare an MCP destructiveHint annotation; "restart-app" implies "terminate" and declares no destructiveHint at all, which the MCP spec reads as destructive by default. See how to fix → Fail
- An AI judge read all 69 captured unit(s) of tool text and found none that tries to manipulate the model reading it.Pass
Capabilities100
- Implements a supported MCP spec version (2025-11-25); the latest is 2026-07-28.Pass
Checks we ran (16)
- Malware scan
- Known CVEs
- Install scripts
- Dependency health
- Source repository
- Provenance
- License
- Maintenance
- Security disclosure
- Schema quality
- Stability
- Tool coverage
- Injection markers
- Destructive annotations
- Judged manipulation
- MCP protocol
Does not apply to this package (2)
- Endpoint SecurityScored on hosted endpoints only. A package has no live endpoint of its own for us to probe.
- Transport & ReachabilityScored on hosted endpoints only. A package has no live endpoint of its own for us to probe.
The score does not measure everything. What the score does not measure →
Every change we have recorded for this component, newest first. Security-relevant changes are always shown. ▲ marks a change for the better, ▼ a change for the worse; unmarked changes are neutral.
- 19 Nov 26 −5
- Stability: pass → unverified ▼ security
- Tool safety: pass → unverified ▼ security
- Stability: pass → fail ▼ security
- Schema quality: 342 → 384 ▼ functional
- Capabilities: pass → unverified ▼ functional
- Tool coverage: 100 → unverified ▼ functional
- First check of Schema quality: unverified functional
- Package version: 0.19.0 → 0.34.0 functional
- 18 Nov 26 +1
- Stability: 0.97 → pass security
- 17 Nov 26 −1
- Stability: pass → 0.97 functional
- 11 Nov 26 +1
- Stability: 0.97 → pass security
- 9 Nov 26 +1
No change was recorded against any check on this day. Stability & Change Management went from 90 to 93. That category is still filling its 30-day observation window: 27 days of observed history at the previous scan, 28 at this one. The score rises as the window fills, whether or not the server changes.
- 7 Nov 26 +1
No change was recorded against any check on this day. Stability & Change Management went from 83 to 87. That category is still filling its 30-day observation window: 25 days of observed history at the previous scan, 26 at this one. The score rises as the window fills, whether or not the server changes.
- 4 Nov 26 −3
- Stability: pass → 0.77 functional
- 29 Oct 26 +1
- Stability: 0.97 → pass security
Diagnostic detail from the automated scan of this channel: what the scanner observed at each step, so you can see exactly where a check passed or failed. It is informational only and never changes the trust score.
Captured 22 Nov 2026 · Analysed npm/@swmansion/argent@0.34.0
Provenance Verified
A signed build attestation was found and verified, binding this exact artifact to the source repository it claims to come from.
| Result | Verified |
|---|---|
| Ecosystem | npm |
| Reason | Verified |
| Discovered via | Registry attestation endpoint |
| Source repo | software-mansion/argent |
| Certificate issuer | https://token.actions.githubusercontent.com |
| Certificate SAN | https://github.com/software-mansion/argent/.github/workflows/publish-npm.yml@refs/heads/main |
| Rekor log index | 3977635893 |
| Predicate type | SLSA build provenance https://slsa.dev/provenance/v1 |
| Subject digest | sha512:da8ac25234d606fcb3f672dced62bb4c9b31c51cd967246bd4efc513db14987ae94d0e60d218b51f7755b6ed2b6fc3a58a9d665dca0814eb8e80bd6b1 |
Background: How many MCP packages publish verified provenance →
Dependencies 177 packages
| Packages resolved | 177 |
|---|---|
| Deprecated | 1 |
| Stale | 63 |
| Tree resolution | Complete |
Background: SBOMs and build attestations, explained →
The tools this component advertises to a client, with an estimated token cost for each. Expand a tool to see its parameters and schema. The per-tool counts are indicative and are not scored directly; the schema's total context footprint is one signal in Schema Quality & AI Usability. A tool's description is untrusted text the model reads on every call, which is what makes this list a security surface and not just an inventory: how tool poisoning works →
react-profiler-fiber-tree ~193
Inspect the React fiber tree and return a JSON representation of the component hierarchy. Use when tracing the ancestry of a library component. Returns a nested JSON tree of fiber nodes with name, tag, actualDuration, selfBaseDuration, and children. Fails if the React DevTools hook is not present or no fiber roots have been committed yet.
| Name | Type | Req | Description |
|---|---|---|---|
| device_id | string | yes | Device ID of the app to inspect. During a recording, use the ID passed to react-profiler-start. |
| filter | string | – | Regex string to filter component names |
| max_depth | integer | – | Maximum tree depth to traverse. Default: 10. Must be greater than 0. |
| port | integer | – | Metro port. Omit it to use this device's port (8081 by default). Ignored for Chromium, whose CDP port is in device_id. Range: 1 to 65535. |
No output schema declared.
No examples provided.
react-profiler-start ~427
Record Hermes CPU samples and React commits during an interaction. On an iOS simulator, Android device, call native-profiler-start in parallel on the same target for a combined report, unless the user declines native profiling. After starting, ask the user to perform the interaction to profile, then call react-profiler-stop. Returns { started_at, startedAtEpochMs, hermes_version, detected_architecture }. On Android, a warning names the apps with native network capture on (started by native-network-logs): their timings in the profile can differ. If a session is already running, returns { already_running: true, mine, owner, … } instead, without clobbering its data, unless it is another tool-server's stale run, has no owner record, or is a lost run of this tool-server: start replaces those. mine: true means this tool-server started the run: if it is yours, call react-profiler-stop. With stale: true it may be another agent's: follow the note. Otherwise, BEFORE OVERTAKING with { force: true }, ask the user for approval first, as how_to_reclaim explains. A start that replaces a run reports it in took_over. Fails if the Hermes runtime is not reachable or the Metro CDP connection cannot be established.
| Name | Type | Req | Description |
|---|---|---|---|
| device_id | string | yes | Device ID from list-devices. For iOS or Android on a shared Metro, use the target's logicalDeviceId if asked to disambiguate. |
| force | boolean | – | Take over an active profiling session even when it is owned by another tool-server and still fresh. Set to true only when you know the prior owner is gone. Default: false. |
| port | integer | – | Metro port. Omit it to use this device's port (8081 by default). Ignored for Chromium, whose CDP port is in device_id. Range: 1 to 65535. |
| sample_interval_us | integer | – | CPU sampling interval in microseconds. Default: 100. Must be greater than 0. |
No output schema declared.
No examples provided.
react-profiler-status ~136
Check whether a React recording is running, saved, lost, or taken over. Use after an interruption to decide whether to collect the recording or start again. If the runtime is unavailable, session_status is no_react_runtime and note gives the reason.
| Name | Type | Req | Description |
|---|---|---|---|
| device_id | string | yes | Device ID of the app to inspect. During a recording, use the ID passed to react-profiler-start. |
| port | integer | – | Metro port. Omit it to use this device's port (8081 by default). Ignored for Chromium, whose CDP port is in device_id. Range: 1 to 65535. |
No output schema declared.
No examples provided.
react-profiler-stop ~254
Stop CPU profiling and collect the cpuProfile + React commit tree. Stores the results for react-profiler-analyze, profiler-commit-query and profiler-cpu-query. Call react-profiler-start first, then exercise the app, then call this. Returns { duration_ms, sample_count, fiber_renders_captured, total_react_commits, hot_commit_indices, react_render_ms, react_render_share_pct, react_render_peak_1s_pct }. If some fibers' display names could not be resolved (typically modals, tooltips or animations that unmounted before stop), it also returns { unattributed_ms, unattributed_fiber_count, unattributed_commit_count }: the work missing from the per-component breakdown. Per-commit durations stay correct. Fails if no active profiling session exists, another stop is running, the CDP connection was lost during recording, or another session took the run over.
| Name | Type | Req | Description |
|---|---|---|---|
| device_id | string | yes | Device ID passed to react-profiler-start. |
| port | integer | – | Metro port. Omit it to use this device's port (8081 by default). Ignored for Chromium, whose CDP port is in device_id. Range: 1 to 65535. |
No output schema declared.
No examples provided.
reinstall-app ~298
Install or reinstall an app on the device. The previous installation (if any) is uninstalled first so app data and runtime permissions are cleared. Use for a full reinstall after rebuilding, or to start from a clean app state. Returns { reinstalled, bundleId }. Fails if `appPath` does not exist or is not the platform's package, or `bundleId` does not match it (not checked on Vega or a remote simulator); these checks run before anything is uninstalled. On Android it also fails, without installing, when bundleId is still on the device after the uninstall or that check fails (the app may then be gone), and on a local iOS simulator when the uninstall does not finish within 30s.
| Name | Type | Req | Description |
|---|---|---|---|
| appPath | string | yes | Path to the app bundle. iOS: `.app` directory (e.g. ./build/.../MyApp.app). Android: `.apk` file (e.g. android/app/build/outputs/apk/debug/app-debug.apk). Vega: `.vpkg` file. Relative paths are resol… |
| bundleId | string | yes | App identifier, used to uninstall the previous installation. iOS: bundle id. Android: package name. Vega: interactive component app id (e.g. com.example.app.main). |
| udid | string | yes | Target device id from `list-devices` (iOS UDID or Android serial). |
No output schema declared.
No examples provided.
restart-app ~266
Terminate then relaunch an app on iOS, Android, or Vega. Use when you need a clean in-memory state without a full reinstall. The relaunch also refreshes the native-devtools injection (iOS and Apple TV). Returns { restarted, bundleId, networkCapture? }. networkCapture (Android) says whether native network capture followed the launch. Fails if the app is not installed, on a local iOS simulator if terminating it does not finish within 15s, or on a remote simulator if terminating it fails for a reason other than the app not running.
| Name | Type | Req | Description |
|---|---|---|---|
| activity | string | – | Android-only: relaunch a non-launcher Activity (e.g. `.SettingsActivity` or `com.example/com.example.SettingsActivity`). If omitted, the app's default launcher activity is used. |
| bundleId | string | yes | App identifier. iOS: bundle id. Android: package name. Vega: interactive component app id. |
| launchArgs | array | – | Apple simulators and physical iPhones: arguments passed to the app process at launch, e.g. `["-FeatureFlag", "YES"]` to override UserDefaults. |
| udid | string | yes | Target device id from `list-devices` (iOS UDID or Android serial). |
No output schema declared.
No examples provided.
rotate ~146
Set the device orientation. Use when testing layout in a different orientation. Re-run `describe` afterwards: frame coordinates change with the orientation. On an unfolded foldable iOS simulator (the iPhone Duo), the value sets the orientation of the device, not of the UI: `Portrait` gives a landscape UI and `LandscapeLeft` gives a portrait UI. To fold or unfold the device, use `fold`. Returns { orientation }. Fails on a TV target, or if the target device is not booted.
| Name | Type | Req | Description |
|---|---|---|---|
| orientation | string | yes | Target orientation |
| udid | string | yes | Target device id from `list-devices` (iOS UDID or Android serial). |
No output schema declared.
No examples provided.
run-sequence ~909
Execute several device interaction steps in one call. Use when every step is known in advance and you do not need to see the screen between steps (e.g. scrolling several times, typing then pressing enter). If a step depends on the result of an earlier one (e.g. a menu item that appears only after a tap), make separate calls instead. Steps run in order. If a step fails (an error, a tool that is not allowed or not supported on the target, or an await-ui-element that returns success=false), the sequence stops there and later steps do NOT run. Returns { completed, total, steps } with a result or an error for each step that ran, so a stopped sequence returns partial results. One screenshot and element tree are captured after the whole sequence, not per step: call screenshot separately only for a baseline before it or to see an intermediate step. A `{{secret:...}}` in any step skips that capture, so type a secret and submit it in one sequence. On a TV target (Apple TV / Android TV / Vega) use tv-remote for remote presses and keyboard for text. Allowed tools, their args (? = optional) and targets. Positions and distances are 0–1 fractions of the screen (window on Chromium), not pixels; angles are in degrees, durations in ms. gesture-tap: x, y, clickCount? [ios/android/chromium] gesture-swipe: fromX, fromY, toX, toY, durationMs?, momentum? [ios/android] gesture-scroll: x, y, deltaX?, deltaY? (signed; positive scrolls down), durationMs? [chromium] gesture-drag: fromX, fromY, toX, toY, durationMs?, momentum? [chromium] gesture-custom: events: [{ type: "Down"|"Move"|"Up", x, y, x2?, y2?, delayMs? }], interpolate? [ios/android] gesture-pinch: centerX, centerY, startDistance, endDistance, endCenterX?, endCenterY?, angle?, durationMs? [ios/android] gesture-rotate: centerX, centerY, startAngle, endAngle, radius (or radiusX + radiusY), durationMs? [ios/android] button: button: "home"|"back"|"power"|"volumeUp"|"volumeDown"|"appSwitch"|"actionButton" [ios/an…
| Name | Type | Req | Description |
|---|---|---|---|
| steps | array | yes | Ordered list of interaction steps to execute sequentially |
| udid | string | yes | Target device id from `list-devices` (iOS UDID, Android serial, Vega serial, or Chromium id), shared across all steps. |
No output schema declared.
No examples provided.
screen-recording-start ~377
Start recording the device screen to a video file (h264 mp4, 30fps at the device's native resolution). The recording keeps running across other tool calls (every result carries a reminder) until `screen-recording-stop` is called or timeLimitSeconds elapses. Immediately after starting, set yourself a reminder/wakeup for the expected end of the recording so it is never left running. Use when the user wants a video of an interaction, animation, or app behavior; for a single still frame use `screenshot` instead. Returns { status: "recording", timeLimitSeconds, outputFile }. The video is retrieved later by `screen-recording-stop`, not by reading outputFile directly. Fails if a recording is already running on the device, the device is not booted, ffmpeg is not installed (local devices only), or the platform cannot be recorded (tvOS, Chromium and Vega are unsupported).
| Name | Type | Req | Description |
|---|---|---|---|
| showTouches | boolean | – | Default true. Draw an on-screen marker for every tap, swipe, drag, pinch and rotate into the video, so it shows where each interaction landed. Set false to record the raw screen with no overlay. |
| timeLimitSeconds | integer | – | Auto-stop cap in seconds (default 180). Set it to slightly more than the interaction you plan to capture. Range: 1 to 600. |
| trimStatic | boolean | – | Default true. Collapse stretches where the screen does not change: the first second of each still stretch is kept, then unchanged frames are dropped until something moves again, so a long recording w… |
| udid | string | yes | Target device id from `list-devices` (iOS Simulator UDID or Android serial). |
No output schema declared.
No examples provided.
screen-recording-stop ~254
Stop the screen recording started by `screen-recording-start` and retrieve the mp4 video. Also retrieves the video when the recording already ended on its own (time limit reached, capture process died); call it even after the cap fired. On a remote (`remote:`) simulator it downloads the mp4. Use when the interaction being captured is finished, or a tool-result note reminds you a recording is still running. Returns { video, durationMs, wallClockMs?, trimmedMs?, panelSwitches?, warning? }; video is a downloadable artifact materialized to a local path. When static-frame trimming removed dead air, durationMs is the trimmed video length and wallClockMs/trimmedMs report the real duration and how much was cut. On a foldable simulator the recording follows the panel that shows the UI: panelSwitches counts the panel changes, the video keeps the size of the first panel, and warning is set when argent could not find that panel during the recording. Fails if no recording (running or finished-but-unretrieved) exists for the given udid.
| Name | Type | Req | Description |
|---|---|---|---|
| udid | string | yes | Target device id from `list-devices` (iOS Simulator UDID or Android serial). |
No output schema declared.
No examples provided.
screenshot ~385
Capture a PNG of the device screen (iOS simulator or physical device, Android emulator, Apple TV simulator, Vega, or Chromium app). Returns { image }, the PNG file. Use when you need a baseline image before an interaction or the screen state after a delay. Gestures, key presses and app launches already attach a screenshot to their result. Fails if the capture backend (simulator-server, emulator, Chromium CDP) cannot be reached for this device.
| Name | Type | Req | Description |
|---|---|---|---|
| downscaler | string | – | Downscaling algorithm when scale<1 on Chromium. Defaults to lanczos3 (highest quality). Ignored on every other platform. |
| includeImageInContext | boolean | – | Set false only when capturing a full-resolution PNG (scale: 1.0) to save as a baseline/current for screenshot-diff. The file is still written, but the image bytes are not attached to the agent contex… |
| rotation | string | – | Orientation override for the screenshot (on Chromium it rotates the captured image). On Android the capture already follows the device's rotation. Ignored on physical iPhones. |
| scale | number | – | Scale factor. Defaults to ARGENT_SCREENSHOT_SCALE env var, or 0.25 if unset for iOS/Android. On a local Apple TV simulator, an Android TV or Vega, with neither set, the capture is downscaled to a 576… |
| udid | string | yes | Target device id from `list-devices` (iOS UDID, Android serial, Apple TV UDID, Vega serial, or Chromium id). |
No output schema declared.
No examples provided.
screenshot-diff ~407
Compare two PNG screenshots and return a compact visual-diff summary. Each side is a saved PNG or a live capture from the device. baselinePath + captureCurrent is the common visual-regression flow. Use when stable before/after screenshots exist and the expected result is pixel-visible: layout, spacing, color, typography, image/icon rendering, clipping, overflow, or text rendering. Physical iPhones: live captures are device-wide and need no registered app. Keep baselines per device model; different aspect ratios fail as a dimension mismatch. Returns { summary, diffPath, contextDiffPath }. The summary uses normalized [0,1] screen locations matching describe coordinates; diffPath is the full-size diff image and contextDiffPath is a downscaled image for MCP/agent display. Ignores the fixed top status-bar band for both pixel and OCR text comparisons. Fails if the input sources are invalid, PNG files cannot be read, outputDir cannot be written, or the simulator-server / emulator backend is not reachable.
| Name | Type | Req | Description |
|---|---|---|---|
| baselinePath | string | – | Path to the saved baseline PNG file. Required unless captureBaseline is true; do not pass both. |
| captureBaseline | boolean | – | Capture the baseline screenshot live from the device, at full resolution, before diffing. Cannot be combined with captureCurrent. |
| captureCurrent | boolean | – | Capture the current screenshot live from the device, at full resolution, before diffing. Cannot be combined with captureBaseline. |
| currentPath | string | – | Path to the saved current PNG file. Required unless captureCurrent is true; do not pass both. |
| outputDir | string | – | Directory where diff artifacts should be written (default: a temp directory). The diff images are returned in the result either way. |
| rotation | string | – | Orientation override for live baseline/current captures. Ignored on physical iPhones. |
| udid | string | yes | Target device id from `list-devices` (iOS UDID or Android serial). |
No output schema declared.
No examples provided.
settings-permissions ~393
Grant or deny an app's runtime permission, or reset it, without navigating the system Settings UI. Use when setting up a test, before the app asks for the permission. iOS simulator: `grant location`/`location-always` needs the app already installed (location authorization is not applied to a bundle id until the app exists). Local simulators enforce this; a remote simulator can't be checked, so install the app there first. Other services can be granted before install. Android: changes the mapped `android.permission.*` runtime permissions (reset also best-effort clears the user-set permission flags). The app must be installed and declare them in its manifest. Some permission changes terminate the app if it is running (system behavior on both platforms), so set permissions before launching, or relaunch after. Returns { action, permission, bundleId, applied, skipped? }: `applied` lists the platform-level services/permissions actually changed; `skipped` (Android) lists mapped permissions the package manager rejected, e.g. ones the manifest doesn't declare. Fails if nothing could be applied.
| Name | Type | Req | Description |
|---|---|---|---|
| action | string | yes | `grant` pre-authorizes the permission, `deny` refuses it, `reset` returns it to the not-yet-asked state so the app prompts on next use. |
| bundleId | string | yes | App to change the permission for; every action, `reset` included, applies to this app only. iOS: bundle id (e.g. com.example.app). Android: package name. |
| permission | string | yes | The permission to change. `notifications` is Android-only; `reminders` is iOS-only; `camera` works on Android, and on iOS only when the simulator's runtime models the service. |
| udid | string | yes | Target device id from `list-devices` (iOS simulator UDID or Android serial). |
No output schema declared.
No examples provided.
shake ~242
Send a shake gesture to the device (iOS simulator or Android emulator). Use when you need to trigger anything bound to the shake gesture: iOS's "Undo Typing" / "Redo Typing" prompt in a text field, React Native's developer menu in a debug build, or an app's own shake-to-report handler. iOS delivers one discrete shake per gesture (the same motion event a physical device raises). Android has no OS-level shake event: the accelerometer is driven through a burst of hard direction changes so app-side detectors fire, and the resting orientation is restored afterwards. Returns { shaken: true, count }. Works on local and remote (sim-remote) iOS simulators. Fails on a physical device or a TV: only phone/tablet simulators and emulators are supported.
| Name | Type | Req | Description |
|---|---|---|---|
| count | integer | – | How many shake gestures to deliver back-to-back (default 1). Raise it for apps whose shake detector needs sustained motion before it fires. Range: 1 to 10. |
| udid | string | yes | Target device id from `list-devices` (iOS simulator UDID or Android emulator serial). |
No output schema declared.
No examples provided.
stop-all-simulator-servers ~590
Stop the services a device owns - simulator-server processes (iOS + Android), native devtools, the iOS accessibility service, TV-control daemons, Chromium CDP sessions, screen recordings, native profiler sessions, Android native network capture (started by native-network-logs), and JS-runtime debugger sessions along with the network inspectors and React profiler sessions that ride on them - freeing their spawned processes, sockets and ports. Call this when the user asks to clean up or ends the session, or when a skill tells you to recycle a device's services mid-flow - not just because a task finished, since the next task reuses these services and would start cold again. To end only the native network capture of one app, call native-network-logs with stop: true instead. PASS `devices` with the device ids this session used: one tool-server serves every agent, subagent and CLI call using this argent install, and an unscoped call tears down THEIR devices too. Omit `devices` only when a machine-wide cleanup is what you actually want. An EMPTY array stops nothing; it does not ask for the machine-wide sweep. A JS-runtime debugger session is keyed by the id you called `debugger-connect` with. On a Metro serving two or more devices that id is not a udid or serial - connect refuses those and tells you to re-target with the `logicalDeviceId` it returns - so a scope built from `list-devices` ids cannot reach that session. Pass any such `logicalDeviceId` in `devices` ALONGSIDE the device id. Returns { stopped }: the URNs of the live services that were shut down; empty means nothing was still running. { unmatched }, present only when a supplied id owns no service here, lists those ids, so a typo or a device NAME passed instead of an id does not read as a clean machine. It is NOT proof the id is wrong: a Vega device you only booted and drove with the remote registers no service and lands here, as does any device this session never started anything on. { left_running } lists the…
| Name | Type | Req | Description |
|---|---|---|---|
| devices | array | – | Ids of the devices THIS session used (iOS UDID / Android serial / Chromium id), to scope the teardown to. |
No output schema declared.
No examples provided.
stop-metro ~206
Stop the Metro bundler process listening on `port`. Use when the user asks to stop Metro or when Metro must be restarted - do not offer it when a task wraps up, since the next task would wait on a cold Metro start. Returns { stopped, port, pids }: stopped=true when it sent SIGTERM to every listening pid it found (or the pid had already exited), stopped=false when it found none. Fails if the port lookup cannot run or times out, if it finds a listener but not the process that owns it, or if it cannot signal a listener it found (e.g. EPERM). On macOS another user's listener is detected only when no other listener on the port is visible, so it can survive stopped=true. This is DESTRUCTIVE: always ask the user for confirmation before calling this tool.
| Name | Type | Req | Description |
|---|---|---|---|
| port | integer | – | TCP port Metro is listening on. Default: 8081. Range: 1 to 65535. |
No output schema declared.
No examples provided.
stop-simulator-server ~236
Stop the transport session for a specific device (iOS simulator / Android: simulator-server process; physical iOS: the on-device runner; Chromium: CDP WebSocket) and free its resources; on a TV target it also reaps that device's TV-control daemons. Use to restart a wedged transport, or when the user is done with one device but wants to keep others running - not just because a task on it finished, since the next task on that device would start cold again. On iOS / Android / TV it deliberately leaves this device's native-devtools, accessibility, profiler and debugger services running - to drain those as well, use stop-all-simulator-servers with `devices`. On CHROMIUM, stopping the transport also ends the JS-runtime debugger session and its captured console history - reconnect with debugger-connect afterwards. Returns { stopped, udid }; `stopped` is false, not an error, when no session is open for the id.
| Name | Type | Req | Description |
|---|---|---|---|
| udid | string | yes | Target device id (iOS UDID, Android serial, or Chromium id) whose transport session to stop |
No output schema declared.
No examples provided.
tv-remote ~343
Press a TV remote / D-pad button (or a whole path of them) on a TV device: Apple TV (tvOS), Android TV (leanback), or Vega (Fire TV). A TV is navigated with a directional remote, not touch. Use when you need to move focus or press a key on a TV, instead of gesture-tap/swipe (which do not apply on a TV). Read the screen with `describe` before and after to see where focus landed. Returns { pressed, count }. Fails on a target that is not a TV, and on the Apple TV simulator for rewind, fastForward, next, previous, volumeUp, volumeDown and mute (its HID stack ignores them); every other button works on all three platforms.
| Name | Type | Req | Description |
|---|---|---|---|
| button | – | yes | A single TV-remote button, or a path of them run in one call. Move focus with up/down/left/right, confirm with select, go back with back/menu, exit with home; the media and volume buttons press the m… |
| repeat | integer | – | Repeat the whole `button` value this many times (default 1). Compact for long same-button runs, e.g. { button: "down", repeat: 12 }. Range: 1 to 50. |
| udid | string | yes | Target TV device id from `list-devices` (Apple TV, Android TV, or Vega). |
No output schema declared.
No examples provided.
update-argent ~146
Apply a pending Argent update. Call this only when the user has explicitly consented in this conversation to updating Argent, e.g. after an update notification reports a new version. Returns { message } with the update status, version info, and whether the tool server restarts. Fails if no update is available or an update is already in progress.
| Name | Type | Req | Description |
|---|---|---|---|
| target | string | – | Which install to update. 'auto' (default) updates the install serving this session: the global PATH install or this project's local devDependency, whichever this server runs from. Pass 'global' / 'lo… |
No output schema declared.
No examples provided.
view-network-logs ~342
Retrieve captured network (HTTP) requests from the running app. Returns a paginated list of requests with method, URL, status, resource type, size, and duration. A size with "~" is estimated from the body's length in characters; on React Native, "→ <url>" is the final URL after a redirect. Pass an entry's requestId to view-network-request-details for full details. On React Native (iOS / Android / Vega) an interceptor injected into the JS runtime records each request once, from XMLHttpRequest (axios included), React Native's fetch, Expo's fetch, and, in a Metro development build, a fetch library that sends through React Native's network module itself, such as react-native-fetch-api. Capture starts at the first network-tool call; a reload clears it. On Chromium it reads the CDP Network domain of the active tab (all request types). Use when inspecting outbound HTTP traffic or debugging API calls in the running app. Fails if the app is not connected (RN) or the device is not reachable (Chromium).
| Name | Type | Req | Description |
|---|---|---|---|
| device_id | string | yes | Device id from list-devices (iOS simulator UDID or Android serial), the same id used with debugger-connect. |
| pageIndex | – | – | Page index (0-based) or "latest" for the most recent page. Each page contains up to 50 entries. Default: "latest". Minimum: 0. |
| port | integer | – | Metro port. Omit it to use this device's port (8081 by default). Ignored for Chromium, whose CDP port is in device_id. Range: 1 to 65535. |
No output schema declared.
No examples provided.
view-network-request-details ~253
Get full details of a specific network request by its requestId. Returns request/response headers (sensitive headers redacted), status, timing, the request body, and optionally the response body. Request and response bodies over 1000 chars are cut to 1000 under a "[TRUNCATED: ...]" line that gives the full size; a binary response body is omitted. Use when you need headers, body, or timing for a specific request after listing logs. Returns an error message string if the requestId is not found.
| Name | Type | Req | Description |
|---|---|---|---|
| device_id | string | yes | Device id from list-devices (iOS simulator UDID or Android serial), the same id used with debugger-connect. An android- ID needs no debugger connection. |
| includeBody | boolean | – | Whether to include the response body (if captured). Default: true. |
| port | integer | – | Metro port. Omit it to use this device's port (8081 by default). Ignored for Chromium, whose CDP port is in device_id. Range: 1 to 65535. |
| requestId | string | yes | The requestId from view-network-logs, or an android- ID from native-network-logs. |
No output schema declared.
No examples provided.
How do I install the Argent MCP server?
Argent runs locally as an npm package, launched with npx -y @swmansion/argent mcp. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.
npm · @swmansion/argent
claude mcp add software-mansion-argent -- npx -y @swmansion/argent mcp
What is the Argent MCP server?
Argent is an MCP server listed in the public MCP registry as io.github.software-mansion/argent. Drive iOS, Android, HarmonyOS, TV, Mac and Electron/web apps from your coding agent. This page covers its npm package (@swmansion/argent).
Is the Argent MCP server safe to use?
Argent scores 86 out of 100 on VerifyMCP. We found no known CVEs affecting it as of 22 November 2026. It declares no install or post-install scripts. Its build provenance is signed and verified. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.
What tools does the Argent MCP server expose?
Argent exposes 68 tools: list-devices, boot-device, launch-app, restart-app, reinstall-app, and 63 more. Their descriptions and schemas cost roughly 25,978 tokens of context every time the server is loaded.
Is the Argent MCP server still maintained?
Argent is still listed as active in the MCP registry. We last reached this channel on 22 November 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.
What licence is the Argent MCP server under?
Argent declares the Apache-2.0 licence, which is OSI-approved. That covers the source only, and says nothing about the cost of any service it calls.
What did VerifyMCP test on the Argent MCP server?
VerifyMCP ran 16 checks on Argent, covering Malware scan, Known CVEs, Install scripts, Dependency health and 12 more. Endpoint Security and Transport & Reachability do not apply to an npm package, so they were not tested. The score does not measure everything, and the scoring documentation lists what it leaves out.