Skip to content
verify mcp Beta VerifyMCP is currently in beta. If you notice any issues, get in touch and we’ll put it right.

Argent

NPM · @SWMANSION/ARGENT

Drive iOS, Android, HarmonyOS, TV, Mac and Electron/web apps from your coding agent

Available components

−5 this week 86 Trust /100

Last scanned (15 hours ago)

Scores higher than 79% of npm servers

Category Browser Automation

Trust breakdown (7 categories)

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
What we tested

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 →

Changelog

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
Diagnostics

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 →

MCP tools · 68 exposed · ~25,978 tokens

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 →

Tool Tokens
await-screen-idle ~216

Block until the screen has rendered content and stopped changing, or the timeout elapses. It polls the same element tree as `describe`. Use when a launch or navigation must finish rendering before a screenshot or tap. Returns { settled, waitedMs, polls, note? }. If the screen never goes still before timeoutMs, it returns settled=false instead of failing.

NameTypeReqDescription
minStableMsinteger–The screen must hold the same content for at least this long to count as settled (default 250). Range: 0 to 10000.
pollIntervalMsinteger–How often to re-read the tree (default 200). Range: 50 to 5000.
timeoutMsinteger–Max time to wait for the screen to settle before giving up (default 3000). Maximum: 120000. Must be greater than 0.
udidstringyesTarget device id from `list-devices` (iOS UDID, Android serial, or Chromium id).

No output schema declared.

No examples provided.

await-ui-element ~529

Block until a UI element reaches an expected state or the timeout elapses, so you don't have to poll screenshot/describe yourself. It polls the same tree as `describe`. Use when the next step needs an element that appears or disappears asynchronously, e.g. after a tap or navigation, or before tapping something that loads late. Returns { success: boolean, elapsed: number, note?, cause? }. If the condition does not hold by the timeout, it returns success=false, and that is not always a verdict on the condition: `cause` says which it was: `unmet` (the tree was read and the condition was false there), `unreadable` (no trustworthy read, so nothing was judged) or `cancelled`. `note` describes what was seen. Only `unmet` licenses rewriting the check.

NameTypeReqDescription
bundleIdstring–iOS app bundle id, passed to the describe fallback (see `describe`). Ignored on every other platform, including physical iOS.
conditionstringyesWhat to wait for. `exists`: the selector matches an element anywhere in the tree. `visible`: it matches one with a non-zero on-screen frame. `hidden`: it matches nothing, or only a zero-area element…
expectedTextstring–Required for condition `text`: the string the checked match must contain or equal (see `textMatch`).
pollIntervalMsinteger–How often to re-check the tree (default 400). Range: 50 to 5000.
selectorobjectyesElement to match. Give at least one field; every given field must match.
textMatchstring–For condition `text`: how expectedText is compared. `contains` (default) is a case-insensitive substring; `equals` is a case-insensitive full-string match.
timeoutMsinteger–Max time to wait for the condition before giving up (default 5000). Maximum: 120000. Must be greater than 0.
udidstringyesTarget device id from `list-devices` (iOS UDID, Android serial, Vega serial, or Chromium id).

No output schema declared.

No examples provided.

boot-device ~881

Start an iOS simulator, launch an Android emulator, start a Vega (Fire TV) Virtual Device, or spawn an Electron app and wait until it is ready to accept interactions. Pass exactly one of 'udid', 'avdName', 'vvdImage', or 'electronAppPath'; it picks the platform. Use when starting a session, once you have picked a target. Returns a tagged payload: { platform: 'ios', udid, booted, note? } or { platform: 'android', serial, avdName, booted, note? } or { platform: 'vega', serial, vvdImage, booted } or { platform: 'chromium', id, port, pid, booted } (an Electron app boots as a Chromium/CDP device). On Android the tool hot-boots from the AVD's default_boot snapshot when usable and falls back to a slower cold boot otherwise. Vega starts the single SDK-managed VVD (~10s). If an Android/Electron boot stage fails, the tool terminates the device it spawned so the next retry starts clean.

NameTypeReqDescription
avdNamestring–Android: AVD name to launch a new emulator from (from `list-devices` → `avds[].name`).
bootTimeoutMsinteger–Android/Vega: overall budget for the boot sequence. Default 480000 (8 min) on Android, 120000 (2 min) on Vega. Range: 30000 to 900000.
electronAppPathstring–Electron: path to the Electron app to launch. Either a packaged .app bundle / executable, or a project directory whose package.json points the Electron binary at the entry script.
electronArgsarray–Electron-only: extra CLI arguments forwarded to the Electron binary after the app path.
electronPortinteger–Electron-only: CDP remote-debugging port to expose. Defaults to a free port; the resulting device id is `chromium-cdp-<port>`. Range: 1024 to 65535.
forceboolean–Shut down and re-boot the device even if already running. Ignored for Electron, which always spawns a fresh process.
guestVulkanboolean–Android on a macOS host only: boot the emulator with guest Vulkan. Defaults to false: guest Vulkan crashes Chrome's GPU process, which breaks Chrome and Custom Tab sign-in flows. Without it apps find…
headlessboolean–iOS only: boot the simulator core WITHOUT opening its GUI window (Simulator.app, or Device Hub on Xcode 27+). The device still streams via simulator-server. The `ARGENT_SIMULATOR_NO_WINDOW` env var (…
soundboolean–Android only (argent never mutes other platforms): boot the emulator with audio output enabled. Defaults to false, so several agent-driven emulators don't all play sound on the host; pass `true` when…
udidstring–iOS: simulator UDID to boot (from `list-devices`).
vvdImagestring–Vega (Fire TV): VVD image to boot: the `vvdImage` of a Vega device from `list-devices` (e.g. `tv`).

No output schema declared.

No examples provided.

button ~154

Press a device hardware button (iOS simulator or physical device, Android emulator or device). Supported buttons depend on the platform: buttons not present on the target (e.g. 'back' on iOS, 'actionButton' on Android, 'power' or 'appSwitch' on a physical iPhone) are rejected with a clear error. Use when you need to trigger hardware button events. Returns { pressed: buttonName }. Fails if the device backend is not reachable: the simulator-server for iOS, or `adb` for Android.

NameTypeReqDescription
buttonstringyesHardware button to press
udidstringyesTarget device id from `list-devices` (iOS UDID or Android serial).

No output schema declared.

No examples provided.

chromium-cookies ~343

Read and write cookies of a Chromium (CDP) app, HttpOnly cookies included. `set` needs `name`, `value`, and `url` or `domain`; `delete` needs `name` and `url` or `domain`. Use when seeding an authenticated session before an e2e test (set the session cookie, then navigate) or asserting cookie state after one. Returns { cookies, count } for get, or a small status object ({ set } / { deleted } / { cleared }) otherwise. Fails if the device is not a Chromium (CDP) device, or a param that `set` / `delete` needs is missing.

NameTypeReqDescription
actionstringyesget: read cookies. set: create/update a cookie. delete: remove matching cookies. clear: remove ALL browser cookies.
domainstring–set/delete: scope the cookie by domain.
expiresnumber–set: expiry as a Unix timestamp (seconds). Omit for a session cookie.
httpOnlyboolean–set: mark HttpOnly.
namestring–Cookie name.
pathstring–set/delete: cookie path (default /).
sameSitestring–set: SameSite policy.
secureboolean–set: mark Secure.
udidstringyesChromium device id from `list-devices` (e.g. `chromium-cdp-9222`).
urlstring–get: restrict to this URL (defaults to the active page). set/delete: scope the cookie by URL.
valuestring–Cookie value.

No output schema declared.

No examples provided.

chromium-tabs ~344

List and switch the tabs / windows of a Chromium (CDP) app (an Electron app's BrowserWindows or a Chromium browser's tabs), and open or close them. The active tab is what describe / gesture-tap / screenshot / debugger-evaluate / open-url all operate on, so select a tab before driving it. tabIds (`t1`, `t2`, …) are stable for the session and never reused. Use when an app exposes multiple windows or tabs and you need to inspect or drive one other than the current page, or to open/close a page during an e2e test. Returns { tabs: [{ tabId, targetId, title, url, active, label? }] } after the action. Fails if the device is not a Chromium (CDP) device, or the requested tabId/label no longer matches a live tab.

NameTypeReqDescription
actionstringyeslist: enumerate tabs/windows. select: make a tab active. new: open a tab and activate it. close: close a tab; if it was active, another live tab becomes active.
labelstring–`new` only: a memorable label usable interchangeably with the tabId.
tabstring–Target tab for `select` (required) / `close`: a tabId like `t2` or a label. `close` defaults to the active tab.
udidstringyesChromium device id from `list-devices` (e.g. `chromium-cdp-9222`).
urlstring–`new` only: URL to open (defaults to about:blank).

No output schema declared.

No examples provided.

debugger-component-tree ~506

Fetch the current screen of a running React Native app as a compact tree of React components. Each on-screen component is listed with its name, its text (or accessibilityLabel when it has no text), [testID=…] and normalized (tap: x,y) centre in [0,1] (fractions of the screen, not pixels). Full-screen transparent wrappers and implementation-detail components are pruned. Wrappers the size of their parent are pruned too, so a missing name proves nothing: check with debugger-inspect-element. Use when you need what describe does not show: a testID for a selector (on Android, describe omits the testID of a view that has no label and is not tappable), or the component names and nesting behind the screen before editing code. To find what to tap, use describe. On an iOS simulator with a landscape UI (a rotated device, or an unfolded foldable), the tap coordinates are on the screen's axes, which the gesture tools use. Returns a compact text tree; call again after navigation. Fails if the Metro debugger is not connected.

NameTypeReqDescription
device_idstringyesDevice id from list-devices: the SAME id you passed to debugger-connect (iOS simulator UDID or Android serial).
includeSkippedboolean–When true, appends a summary of all filtered components: total fiber count and skip counts by component name and by filter. Default: false.
maxNodesnumber–Maximum total nodes to include. When exceeded, intermediate single-child wrapper chains are collapsed to preserve both root structure and leaf elements. Default: no limit.
onScreenOnlyboolean–When true, only components visible on screen are returned. Set to false to include all mounted components, including those scrolled off-screen, e.g. to understand the full page structure. Default: tr…
portinteger–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.
udidstring–iOS simulator UDID from list-devices. Pass it when device_id is a logicalDeviceId (two or more devices share one Metro), so that the tap coordinates of a landscape UI are on the screen's axes. Give t…

No output schema declared.

No examples provided.

debugger-connect ~300

Connect to a JS runtime CDP debugger. iOS / Android / Vega: connects to Metro's CDP endpoint on the given port. Chromium: re-uses the page CDP session opened by boot-device. Returns connection info including port, projectRoot (empty on Chromium and on legacy Metro, e.g. Vega), deviceName, appName, logicalDeviceId, and isNewDebugger, plus a note naming the kept log file when the previous session ended because the app crashed. If already connected, returns the existing connection. Use when starting a debug session or before calling other debugger-* tools. Fails if the runtime is unreachable (Metro down, or Chromium CDP terminated).

NameTypeReqDescription
device_idstringyesDevice id from list-devices: iOS simulator UDID, Android serial, Vega serial (amazon-...), or Chromium device id (chromium-cdp-<port>). Pass this SAME id as device_id to every subsequent debugger-* c…
portinteger–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.

debugger-evaluate ~247

Execute arbitrary JavaScript in the app's JS runtime via CDP: Hermes on iOS / Android / Vega, V8 on Chromium. Returns the result serialized by value as JSON, with deviceName, appName, and logicalDeviceId. Use when you need to read app state, including localStorage and sessionStorage on Chromium, call app functions, or test logic at runtime. Cyclic objects, which many RN runtime values are (fiber nodes, navigation refs, global), fail with a serialization error. Fails if the expression throws or the runtime is not connected.

NameTypeReqDescription
device_idstringyesDevice id from list-devices: the SAME id you passed to debugger-connect (iOS simulator UDID, Android serial, Vega serial, or Chromium device id). The logicalDeviceId debugger-connect returns also res…
expressionstringyesJavaScript expression to evaluate in the app runtime
portinteger–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.

debugger-inspect-element ~397

Inspect the React component hierarchy at a screen coordinate (x, y). Returns components from the tapped element upward through its parent hierarchy, each with its source file:line and a code fragment. The first items (lowest indices) are the most specific: the exact component under the tap point and its direct parents. Higher indices are broader context (page, navigator). Use when you need the source file and line for a component at a tap coordinate. Fails if the app is not connected or the coordinate is outside the screen.

NameTypeReqDescription
contextLinesnumber–Lines of source context to include around the component definition. Default: 3.
device_idstringyesDevice id from list-devices: the SAME id you passed to debugger-connect (iOS simulator UDID or Android serial).
includeSkippedboolean–When true, items that would normally be filtered are kept in the response with skipped=true and a skipReason. Default: false.
maxItemsnumber–Maximum number of hierarchy items to return, most specific first. The default covers all app-specific code; increase to 70+ to also see the navigation/screen structure. Default: 35.
portinteger–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.
resolveSourceMapsboolean–When true, resolves bundled frame locations to original source files via Metro symbolication and includes a code fragment. When false, returns the raw bundled frame info (file, line, column) without…
xnumberyesX coordinate on device screen, in logical pixels (not normalized 0–1)
ynumberyesY coordinate on device screen, in logical pixels (not normalized 0–1)

No output schema declared.

No examples provided.

debugger-log-registry ~376

Get a summary of all console logs captured from the app's JS runtime. Returns the log file path, entry counts by level, and message clusters (grouped by similarity). Works against Hermes (iOS / Android / Vega) and V8 (Chromium). Use when investigating warnings, errors, or unexpected output. Call this first for an overview, then Grep the returned file for details; never Read it whole, it can hold 50,000 lines. Returns empty stats if no log data has been captured yet, but check { note }, which is present only when the stats are empty BECAUSE the previous debugger session ended: torn down (its log file deleted) or cut off by an app crash (its log file kept at the path the note names). Without that note, empty means the app has logged nothing. When the debugger cannot be reached, it returns { status: "not_connected", reason, detail, guidance } with NO log file. Follow the guidance (do not retry in a loop). Grep a log file in this state only when a { note } names one kept from before the app crashed. A "connected" result's stats may come from a session whose socket has since died. Use debugger-status, not this tool, to judge debugger health.

NameTypeReqDescription
device_idstringyesDevice id from list-devices: the SAME id you passed to debugger-connect (iOS simulator UDID, Android serial, Vega serial, or Chromium device id). The logicalDeviceId debugger-connect returns also res…
portinteger–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.

debugger-reload-metro ~156

Restart the Metro JS bundle in the connected React Native app without restarting the native process. Use when you want to apply code changes or reset JS state. Returns { reloaded, port, method (cdp or http), deviceName, appName, logicalDeviceId }. Fails if Metro is not running on the given port.

NameTypeReqDescription
device_idstringyesDevice id from list-devices: the SAME id you passed to debugger-connect (iOS simulator UDID or Android serial).
portinteger–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.

debugger-status ~298

Get JS runtime debugger connection status and diagnostic info. Use when you need to verify connectivity before using other debugger tools. Never fails when the runtime is simply unreachable: it returns a discriminated result instead: - { status: "connected", ... } with port, projectRoot (empty on Chromium and on legacy Metro, e.g. Vega), deviceName, appName, logicalDeviceId (absent on Vega), isNewDebugger (false on the legacy inspector), connected flag, loadedScripts count, and sourceMapReady (always true: waits for pending source maps before returning). - { status: "not_connected", connected: false, reason, detail, guidance } (port omitted on Chromium), with reason one of metro_not_running, no_app_connected, device_mismatch, cdp_unreachable, runtime_unresponsive, stale_connection, reconnecting. Follow the guidance field; do not retry in a loop.

NameTypeReqDescription
device_idstringyesDevice id from list-devices: the SAME id you passed to debugger-connect (iOS simulator UDID, Android serial, Vega serial, or Chromium device id). The logicalDeviceId debugger-connect returns also res…
portinteger–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.

describe ~581

Get the accessibility / DOM element tree for the current screen. Use when you need element frames or labels and have no fresh tree: most interaction tools (not tv-remote or shake) already return one after each action, or name an identical earlier one by its tree id. On iOS it reads the AXRuntime accessibility service, so it sees whatever is visible, including system dialogs and permission prompts. While a system dialog is visible it returns the dialog's elements (buttons, text); otherwise the foreground app's. On a physical iOS device it covers only the app started with launch-app; to read system dialogs there, launch-app `com.apple.springboard` first. On Android it reads the tree through Argent's UiAutomation helper, falling back to `uiautomator dump`. On Chromium it walks the DOM over the Chrome DevTools Protocol: every visible element with its ARIA role, accessible name and bounding rect. On Vega (Fire TV) it reads the on-device automation toolkit; `[focused]`/`[selected]` mark the D-pad cursor, which you move with `tv-remote` (Vega has no touch). On a TV target (Apple TV / Android TV: a `list-devices` entry with runtimeKind 'tv') it returns the FOCUSED element and the FOCUSABLE elements instead, since a TV UI has no tap coordinates. Move the highlight with `tv-remote`, then call describe again to confirm where focus landed. Returns `{ description, source }`: `description` renders the tree one line per element, with its role, label/value/id, VoiceOver hint (iOS simulator), interactivity flags and frame. Frames are normalized [0,1] fractions of the screen / window width and height (not pixels); tap an element at the centre of its frame. The tree has no z-order or occlusion information: an element may be covered by an overlay (e.g. a toolbar over list rows), so check a screenshot when a tap lands unexpectedly. Fails if the tree cannot be read, e.g. on a physical iOS device before launch-app has started an app. For UIKit properties of a view (class name, accessib…

NameTypeReqDescription
bundleIdstring–App bundle ID, used as a target hint on iOS when the AX-service returns no elements and describe falls back to native-devtools inspection. If omitted, the fallback auto-detects the frontmost connecte…
udidstringyesTarget device id from `list-devices` (iOS UDID, Android serial, Vega serial, or Chromium id).

No output schema declared.

No examples provided.

dismiss-update ~131

Postpone the Argent update reminder, on every tool-server on this machine, when the user wants it later or tomorrow, wants it stopped, or never wants it (87600 hours). Installs nothing. A memory or a reminder of your own does not change it. Returns { message, remindersResumeAt, persisted, pendingVersion }: tell the user when reminders resume, and if persisted is false, that it did not fully hold.

NameTypeReqDescription
hoursnumberyesHours until reminders resume: 168 is a week, 87600 is the longest, 0 turns them back on. Minimum: 0.

No output schema declared.

No examples provided.

e2e-add-echo ~172

Record an echo step in the e2e test named by `name` + `project_root`. Echo steps print a message when the e2e test is replayed, useful as labels between tool calls. Use when you want to annotate a recorded e2e test with a human-readable label or checkpoint message. Returns { message, stepCount, savedTo }. Fails if that e2e test has no recording in progress.

NameTypeReqDescription
messagestringyesText the echo step prints
namestringyesName of the e2e test being recorded: the one passed to e2e-start-recording.
project_rootstringyesAbsolute path to the project root of the e2e test being recorded: the same value passed to e2e-start-recording.

No output schema declared.

No examples provided.

e2e-add-script ~251

Run a local JavaScript (.mjs, .js) or TypeScript (.mts, .ts) file and record it as a `script:` step in an active e2e test. Use when the user requests a local script in the e2e test, and only then. Call it at the point in the recording where the script must run. Returns `status`, `reason` and `stepCount`. A failed script is not recorded. Check `reason` and the affected state before you retry. Over a link (argent link or ARGENT_TOOLS_URL), the tool refuses the call: record such an e2e test with no link.

NameTypeReqDescription
namestringyesE2E test name passed to e2e-start-recording.
pathstringyesPath to the script file (.mjs, .mts, .js or .ts), relative to the e2e test YAML. For example: "../../scripts/seed-order.mjs".
project_rootstringyesAbsolute project root passed to e2e-start-recording.
timeoutnumber–Time limit in milliseconds. The default is 30000 and the minimum is 100.

No output schema declared.

No examples provided.

e2e-add-step ~308

Execute one MCP tool and record its step in the e2e test named by `name` + `project_root`. Use when recording and you want each action run and captured; the recording must already be open. A coordinate `gesture-tap` records as a portable `tap` selector step; `restart-app` as a `launch`. Returns { message, stepCount, recorded, savedTo }; `recorded`, not the status, says whether a step was appended. Fails with an error, recording nothing. Call recording tools, including `e2e-add-script`, directly.

NameTypeReqDescription
argsstring–Tool arguments as a JSON string, e.g. '{"udid": "ABC", "x": 0.5, "y": 0.3}'. Omit for tools with no arguments.
commandstringyesMCP tool to execute and record, for example "gesture-tap". Do not pass an e2e test directive or a recording tool. Call e2e-add-script directly for a requested script step.
delayMsinteger–Milliseconds to sleep before executing this step during replay. Minimum: 0.
namestringyesName of the e2e test being recorded: the one passed to e2e-start-recording.
project_rootstringyesAbsolute path to the project root of the e2e test being recorded: the same value passed to e2e-start-recording.

No output schema declared.

No examples provided.

e2e-execute ~890

Run a saved YAML e2e test end to end. Use when asked to replay a recorded path, re-run a QA regression, or check that a known journey still passes; for a one-off interaction use the gesture tools instead, and to author an e2e test use e2e-start-recording. Pass exactly one e2e test source: name (under project_root) or e2e_test_path. Returns a per-step report: the first failure stops the run and the rest report as skipped. Over a link (argent link or ARGENT_TOOLS_URL), the client uploads the e2e test with the run: fragments, snapshot baselines and absolute .png/.yaml file arguments its steps use, from under project_root or beside the e2e test file. The run fails if a step needs any other project file (a script: step, a recording tool, a nested e2e-execute not called by name with an absolute project_root, a directory, app or output-directory argument, a file path built from several arguments, a relative path): it stops before the first step and the error lists each such step. An absolute .png/.yaml argument from anywhere else fails only when its step runs. Without a link, every step kind runs.

NameTypeReqDescription
devicestring–Device id to run against (iOS UDID, Android/Vega serial, Chromium id), as list-devices reports it. Auto-detected when omitted, but only when exactly one booted device matches (optionally narrowed by…
e2e_test_filestring–Path to the e2e test .yaml as readable by the tool-server. Internal: the argent client derives it from project_root and name automatically; leave unset.
e2e_test_pathstring–Absolute path to an e2e test .yaml on the client.
namestring–Name of a saved e2e test to run from `<project_root>/.argent/e2e`, or `.argent/flows` when it is not there (e.g. "settings-explore").
platformstring–Restrict auto-detection to this platform when several devices are booted. `ios` selects local simulators only; `ios-remote` selects a remote one. `chromium` with no `device` also makes the runner boo…
prerequisiteAcknowledgedboolean–Set to true to confirm the execution prerequisite has been met. Required when a fragment defines an executionPrerequisite: without it, the call runs no step and returns the prerequisite text. Not a d…
project_rootstringyesAbsolute path to the calling agent's project root: the cwd it is working in, also with e2e_test_path, where the e2e test, its run: siblings, its script: paths and baselines resolve beside the YAML in…
updateBaselinesboolean–Write/refresh screenshot baselines for `snapshot` steps instead of diffing against them, also in the e2e tests that tool: e2e-execute steps run, unless such a step sets updateBaselines itself. Over a…

No output schema declared.

No examples provided.

e2e-finish-recording ~248

Finish recording the e2e test named by `name` + `project_root`, leaving recordings under any other key untouched. Returns { message, path, executionPrerequisite, steps, summary, e2eTestFile, savedTo } - a summary of all recorded steps plus the final YAML. Use when you have added all desired steps and want to finalize the e2e test file. Fails if that e2e test has no recording in progress. A warning e2e-add-step raised on a recorded `await-ui-element` is repeated in `summary` as a `warning:` line right below its step, and `message` counts them by kind. Read them before you convert a wait to `await:`/`assert:` or replay the e2e test. You can still edit the .yaml file directly afterwards to remove or reorder steps.

NameTypeReqDescription
namestringyesName of the e2e test being recorded: the one passed to e2e-start-recording.
project_rootstringyesAbsolute path to the project root of the e2e test being recorded: the same value passed to e2e-start-recording.

No output schema declared.

No examples provided.

e2e-start-recording ~769

Start recording a new e2e test, resetting .argent/e2e/<name>.yaml to an empty e2e test and replacing any existing one. Use when you want to capture a reusable sequence of device interactions for later replay. Load the argent-create-e2e skill first: it holds the polish, audit and replay steps. Not for Apple TV or Android TV. Returns { message, e2eTestFile, savedTo } and optionally { restarted, discardedSteps } if a live recording of the same e2e test was discarded. Without a link, it creates that file and fails if the .argent/e2e/ directory cannot be created or the file cannot be written. Over a link (argent link or ARGENT_TOOLS_URL, also to 127.0.0.1), your client writes the file, and `savedTo` is its path on your computer, or null when that write failed. Several e2e tests can be recorded at once, each keyed by the `name` + `project_root` that every subsequent recording tool repeats, and one recording's steps never land in another's file. After starting, use e2e-add-step to append tool calls. Each step runs LIVE, so give each concurrent recording its own device and pick a name unique to your task. Read each step's `message`: it warns when a recorded await-ui-element did not hold, or may fail once converted to `await:`/`assert:`. A wait nested in a recorded run-sequence gets no such warning, so read its `toolResult`. For a standalone e2e test, record a restart-app of the app under test as the FIRST step (captured as the e2e test's `launch` step); for a reusable fragment, skip that and pass executionPrerequisite instead. Starting over an existing e2e test replaces the whole file, including its `requires:` block; pass `requires` to keep it. Use e2e-add-echo to add labels, and e2e-add-script to run a local .mjs, .mts, .js or .ts file and record it as a `script:` step. Call e2e-finish-recording when done. To remove or reorder steps, edit the .yaml file after e2e-finish-recording, not during the recording: over a link, the next recorded step writes over your edit, and…

NameTypeReqDescription
executionPrerequisitestring–Fragments only: the app/device state assumed on entry (e.g. "Settings app open on General page"). Omit it for a standalone e2e test. restart-app has no chromium support, so a chromium e2e test record…
namestringyesName for this e2e test (e.g. "settings-explore"): letters, digits, underscore and hyphen only.
project_rootstringyesAbsolute path to the project root directory (the directory that contains or should contain `.argent/e2e/`).
requiresobject–The e2e test's `requires:` block: the targets it supports, ANDed. `platform` is one platform or a list (`ios` also covers a remote simulator); `runtimeKind` is `tv` or `mobile`. Omit it for a test th…

No output schema declared.

No examples provided.

fold ~507

Move the hinge of a foldable iOS simulator (the iPhone Duo) or Android emulator (a Pixel Fold or generic foldable AVD) to a `posture` or to an `angle`; pass exactly one of the two. The tool returns when the device accepts input again, so the next tap lands. Use when you must see or test an app in a different posture. `list-devices` marks foldable devices with `foldable: true`. To turn a device without a fold, use `rotate`. All tools use the active panel (the panel that shows the UI), also after a fold made outside argent. The coordinates and the screenshot size change with the panel. Before you tap, read the element tree in this result or run `describe` again. iOS: unfolded, the UI is landscape, as on a rotated iPhone. A fold between two angles that are not closed or open can keep the current panel. To change panels, fold to closed or open. iOS returns { activeScreen, screen: { id, panel, width, height }, posture?, hingeAngle, warning? }. screen names the active panel and gives its size in pixels. posture is set only at 0°, 120° and 180°. warning is set when argent cannot find the active panel, or when the device did not change to the expected panel. Android returns { posture?, deviceState, hingeAngle?, warning? }. deviceState is the state Android committed (CLOSED, HALF_OPENED, OPENED), and posture is the matching preset. A lock screen the fold raises is dismissed; warning is set when it cannot be, or when a lock screen is showing or the screen is off after the fold. On a generic foldable AVD, closed shows only the left part of the display and describe frames do not match tap coordinates; use a Pixel Fold AVD to test the cover panel. Fails if the device is not foldable, is a remote simulator or a physical Android device.

NameTypeReqDescription
anglenumber–The hinge angle in degrees: 0 = closed, 180 = open. Range: 0 to 180.
posturestring–Where to move the hinge: `closed` shows the cover panel, `half-open` and `open` show the inner panel.
udidstringyesTarget device id from `list-devices` (a foldable iOS simulator UDID or Android emulator serial).

No output schema declared.

No examples provided.

gesture-custom ~628

Send a sequence of touch events for complex gestures. Use for: long press, drag-and-drop, custom scroll, pinch (second touch point). For simple taps use the gesture-tap tool. For straight-line scrolling use the gesture-swipe tool. For pinch gestures use gesture-pinch. For rotation gestures use gesture-rotate. All x/y values are fractions of the screen width/height (not pixels). Physical iOS: a Down, any Moves, then an Up; x2/y2 adds a second finger. A plain Down then Up: same point = press-hold, other point = drag (Up delayMs = rest before the move); one Move at the Down point between them = long-press pickup (Move delayMs = hold, Up delayMs = move time). Any other train plays as timed paths, each event waiting its delayMs, and needs the app drawn in portrait. Returns { events: number } with the total count of events dispatched. On physical iOS, reactivated: true = app was re-fronted; re-describe. Fails if the target device is not booted or an event type is invalid. Example long-press at center: [{"type":"Down","x":0.5,"y":0.5},{"type":"Up","x":0.5,"y":0.5,"delayMs":800}] Example pick up a list item with a long press, then drag it up: [{"type":"Down","x":0.5,"y":0.6},{"type":"Move","x":0.5,"y":0.6,"delayMs":800},{"type":"Up","x":0.5,"y":0.3,"delayMs":500}] Example smooth scroll down: [{"type":"Down","x":0.5,"y":0.7}, {"type":"Move","x":0.5,"y":0.6},{"type":"Move","x":0.5,"y":0.5},{"type":"Move","x":0.5,"y":0.4}, {"type":"Up","x":0.5,"y":0.3}] Example pinch-to-zoom (with interpolate:10 for smoothness): events: [{"type":"Down","x":0.4,"y":0.5,"x2":0.6,"y2":0.5}, {"type":"Up","x":0.2,"y":0.5,"x2":0.8,"y2":0.5}] interpolate: 10

NameTypeReqDescription
eventsarrayyesTouch events, dispatched in order.
interpolatenumber–Number of Move events to insert between each pair of consecutive events, linearly interpolating both x,y and x2,y2 and splitting the delay evenly, for a smooth gesture. Default: no interpolation.
udidstringyesTarget device id from `list-devices` (iOS UDID or Android serial).

No output schema declared.

No examples provided.

gesture-drag ~376

Press the left mouse button at a start point, move to an end point, and release: a desktop mouse drag in a Chromium app, with mouse moves interpolated at ~60fps. All positions are fractions of the window width/height (not pixels), the same space as gesture-tap and describe, except that exactly 1.0 lands one pixel inside the window edge. Use for slider thumbs, drag-and-drop, text selection, or draggable UI elements. Dragging never scrolls content on desktop; use gesture-scroll for lists/pages. Chromium only: on iOS/Android use gesture-swipe. Returns { dragged: true, timestampMs }. Fails if the Chromium CDP session is not reachable for the given device.

NameTypeReqDescription
durationMsnumber–Total drag duration in milliseconds (default 300); the button stays down this long. Maximum: 10000.
fromXnumberyesPress x. Range: 0 to 1.
fromYnumberyesPress y. Range: 0 to 1.
momentumboolean–Whether the drag releases with momentum; default true (a constant-speed drag). false decelerates into the release point, so an app that derives a fling from the release velocity (carousels, drag libr…
toXnumberyesRelease x. Range: 0 to 1.
toYnumberyesRelease y. Range: 0 to 1.
udidstringyesTarget Chromium device id from `list-devices` (chromium-cdp-<port>).

No output schema declared.

No examples provided.

gesture-pinch ~407

Execute a pinch-to-zoom gesture by moving two fingers toward or away from a center point to change the scale of on-screen content. Positions and distances are fractions of the screen (0.0–1.0, not pixels), as in gesture-tap. startDistance > endDistance = pinch in (zoom out). startDistance < endDistance = pinch out (zoom in). Typical values: startDistance 0.2, endDistance 0.6 for a zoom-in pinch at screen center. Use when you need to zoom in or out on a map, image, or zoomable view. Returns { pinched: true, timestampMs }. On physical iOS, reactivated: true = app was re-fronted; re-describe. Fails if the simulator-server / emulator backend is not reachable for the given device, or on a physical iPhone when the app is not drawn in portrait.

NameTypeReqDescription
anglenumber–Angle in degrees of the axis the fingers lie on: 0 = horizontal (default), 90 = vertical.
centerXnumberyesCenter of the pinch, horizontal.
centerYnumberyesCenter of the pinch, vertical.
durationMsnumber–Total gesture duration in milliseconds (default 300)
endCenterXnumber–Final horizontal center of the pinch. When set, the centroid drifts linearly from centerX to endCenterX over the gesture (e.g. to keep expanding fingers on-screen near an edge). Omit for a fixed cent…
endCenterYnumber–Final vertical center of the pinch; drifts from centerY like endCenterX. Omit for a fixed center.
endDistancenumberyesFinal distance between the two fingers.
startDistancenumberyesInitial distance between the two fingers.
udidstringyesTarget device id from `list-devices` (iOS UDID or Android serial).

No output schema declared.

No examples provided.

gesture-rotate ~464

Send a two-finger circular arc gesture to rotate on-screen content by a specified angle. Two fingers are placed opposite each other at a fixed radius from the center, then swept from startAngle to endAngle degrees (endAngle > startAngle = clockwise). All positions and radii are normalized 0.0–1.0 (fractions of screen width/height, not pixels), same coordinate space as gesture-tap and gesture-swipe. Typical values: radius 0.15, startAngle 0, endAngle 90 for a 90° clockwise turn. Size the orbit with radius, or with radiusX and radiusY together (the pair overrides radius); one half of the pair alone, or none of the three, is rejected. A single radius traces a physical ellipse on a non-square screen (finger separation varies through the turn); for a physically circular orbit, pass radiusX and radiusY with radiusX·width = radiusY·height. Unlike gesture-pinch which moves fingers linearly to zoom, this orbits fingers in an arc to change orientation. Use when you need to rotate a map, image picker, or any rotateable UI element. Returns { rotated: true, timestampMs }. On physical iOS, reactivated: true = app was re-fronted; re-describe. Fails if the simulator-server / emulator backend is not reachable for the given device, or on a physical iPhone when the app is not drawn in portrait.

NameTypeReqDescription
centerXnumberyesCenter of rotation, horizontal.
centerYnumberyesCenter of rotation, vertical.
durationMsnumber–Total gesture duration in milliseconds (default 300)
endAnglenumberyesEnding angle in degrees.
radiusnumber–Distance from the center to each finger.
radiusXnumber–Finger distance along the horizontal axis, as a fraction of screen WIDTH.
radiusYnumber–Finger distance along the vertical axis, as a fraction of screen HEIGHT.
startAnglenumberyesStarting angle in degrees (0 = right, 90 = down)
udidstringyesTarget device id from `list-devices` (iOS UDID or Android serial).

No output schema declared.

No examples provided.

gesture-scroll ~300

Send mouse-wheel events at a point to scroll content in a Chromium app. The anchor x/y and the deltas are fractions of the window (not pixels); x/y match gesture-tap and describe. Use when content is below/above the fold (describe shows off-screen elements with zero height) or a list needs scrolling. Chromium only: on iOS/Android use gesture-swipe. Returns { scrolled: true, timestampMs }. Fails if the Chromium CDP session is not reachable for the given device.

NameTypeReqDescription
deltaXnumber–Horizontal scroll distance as a fraction of the window width. Positive reveals content to the right; negative scrolls back left.
deltaYnumber–Vertical scroll distance as a fraction of the window height (0.5 = half a window). Positive scrolls down (reveals content below), like rolling a mouse wheel toward you; negative scrolls back up.
durationMsnumber–Spread the scroll over this many milliseconds in wheel-event steps (default 300) so scroll handlers fire progressively.
udidstringyesTarget Chromium device id from `list-devices` (chromium-cdp-<port>).
xnumberyesAnchor x. The wheel events land here; put it over the element you want to scroll. Range: 0 to 1.
ynumberyesAnchor y. Range: 0 to 1.

No output schema declared.

No examples provided.

gesture-swipe ~457

Execute a smooth swipe / drag touch gesture between two points on the device (iOS simulator or physical device, or Android emulator). All from/to positions are fractions of screen width/height (not pixels), same as gesture-tap. Generates interpolated Move events for a natural feel (~60fps). Swipe up (fromY > toY) to scroll content down. Use when you need to scroll a list, dismiss a modal, drag an element, or navigate between pages. Not supported on Chromium; use gesture-scroll there instead. Physical iOS: an edge gesture (back-swipe) needs fromX 0 exactly; durationMs sets drag speed, not time; momentum:false only rests 300ms at the end and does not damp. Returns { swiped: true, timestampMs }. On physical iOS, reactivated: true = app was re-fronted; re-describe. Fails if the simulator-server / emulator backend is not reachable for the given device.

NameTypeReqDescription
durationMsnumber–Total gesture duration in milliseconds (default 300); the finger stays down this long. Below ~32ms the whole travel lands in one or two frames, which the OS flings as hard as any flick. Maximum: 1000…
fromXnumberyesStart x. Range: 0 to 1.
fromYnumberyesStart y. Range: 0 to 1.
momentumboolean–Whether the swipe releases with momentum; default true (a natural flinging swipe). false eases the finger into the end point so the OS applies little to no fling and the content stops where the finge…
toXnumberyesEnd x. Range: 0 to 1.
toYnumberyesEnd y. Range: 0 to 1.
udidstringyesTarget device id from `list-devices` (iOS UDID or Android serial).

No output schema declared.

No examples provided.

gesture-tap ~317

Press the device screen (iOS simulator or physical device, Android emulator, or Chromium app) at a point given as fractions of the screen width and height (not pixels): a Down event followed by an Up event at the same spot. Use when tapping a button, link, or any tappable element on the screen. Returns { tapped: true, timestampMs }. On physical iOS, reactivated: true = app was re-fronted; re-describe. Fails if the device backend (simulator-server / emulator / Chromium CDP) is unreachable. Before tapping, take the coordinates from `describe` on every platform (on Chromium it walks the DOM). More information in `argent-device-interact` skill

NameTypeReqDescription
clickCountinteger–Number of taps/clicks dispatched as ONE gesture inside the OS double-tap window (2 = double-tap / double-click), which two separate tap calls cannot guarantee. On physical iOS, higher counts loop sin…
udidstringyesTarget device id from `list-devices` (iOS UDID, Android serial, or Chromium id).
xnumberyesHorizontal position: 0 = left edge, 1 = right edge. Range: 0 to 1.
ynumberyesVertical position: 0 = top edge, 1 = bottom edge. Range: 0 to 1.

No output schema declared.

No examples provided.

keyboard ~812

Type text or press special keys on the device (iOS simulator, Android emulator or device, Chromium app, Vega Virtual Device, or Apple TV / Android TV) using keyboard events. Use when you need to enter text or trigger a named key such as enter, escape, or arrow keys. On Vega and Apple TV / Android TV, prefer the remote tools for D-pad navigation; focus a text field with `tv-remote` first (e.g. a search or login box), then type into it with keyboard. Tap the field first: with no focus the text is dropped and the call still succeeds. Returns { typed: string, keys: number }. On physical iOS, reactivated: true = app was re-fronted; re-describe. Fails if an unsupported key name is provided (rejected before anything is sent) or if the device's input backend is not reachable. A failure is not rolled back. Un-typeable text is rejected mid-string on the iOS simulator and Chromium, leaving the characters before it in the field (Android, Vega and TV targets check the whole string up front). A transport failure partway also leaves the text already sent. On a retry, read the field's actual contents; do not assume it is unchanged. React Native: after typing, the next tap outside the field may only close the keyboard. If only the keyboard closed, tap again; where the field submits on return, `key: "enter"` avoids the extra tap.

NameTypeReqDescription
delayMsnumber–Delay in ms between key presses (default 50). Ignored on Android phones/tablets, Vega, Apple TV / Android TV and physical iOS, which type the whole string in one shot or at their own cadence.
keystring–Named key to press: enter, escape, backspace, tab, space, arrow-up, arrow-down, arrow-left, arrow-right, f1-f12. Cannot be combined with `text`; to do both, send two `keyboard` steps in one `run-sequ…
textstring–Text to type character by character; handles uppercase, digits and common punctuation. Cannot be combined with `key`: to type and then press a key, send two `keyboard` steps in one `run-sequence` ({…
udidstringyesTarget device id from `list-devices` (iOS UDID, Android serial, Vega serial, or Chromium id).

No output schema declared.

No examples provided.

launch-app ~484

Open an installed app, or confirm the running renderer (Chromium). Use when starting any app; prefer this over tapping home-screen / launcher icons. Also prepares the native-devtools injection before the app starts (iOS and Apple TV). Returns { launched, bundleId, note?, networkCapture? }. Fails if the app is not installed on the target device (iOS / Android). On a physical iPhone this registers the app every other tool acts on; com.apple.springboard and com.apple.Spotlight register without launching. note warns when runner signing is not ready. networkCapture (Android) says whether native network capture followed the launch. For Chromium, the app is already running behind a CDP port; this call simply refreshes the cached viewport and acknowledges the bundleId tag. To change the visible route, use `open-url`. Common iOS bundle ids: com.apple.MobileSMS, com.apple.mobilesafari, com.apple.Preferences, com.apple.Maps, com.apple.mobileslideshow (Photos), com.apple.MobileAddressBook Common Android packages: com.android.settings, com.android.chrome, com.google.android.apps.maps, com.google.android.gm, com.android.vending, com.google.android.dialer, com.google.android.apps.messaging

NameTypeReqDescription
activitystring–Android-only: fully-qualified Activity name (e.g. `.MainActivity` or `com.example/com.example.MainActivity`). If omitted, the app's default launcher activity is used.
bundleIdstringyesApp identifier. iOS: bundle id. Android: package name from build.gradle `applicationId`. Vega (Fire TV): the interactive component app id from manifest.toml (e.g. com.example.app.main). Chromium: any…
launchArgsarray–Apple simulators and physical iPhones: arguments passed to the app process at launch, e.g. `["-FeatureFlag", "YES"]` to override UserDefaults. A running app is relaunched to apply them. The call fail…
udidstringyesTarget device id from `list-devices` (iOS UDID, Android serial, or Chromium id).

No output schema declared.

No examples provided.

list-devices ~560

List iOS simulators, Android emulators, connected physical Android devices, running Chromium apps, and Vega (Fire TV) devices in one place. Call this at the start of a session to pick a target id ('udid' for iOS entries, 'serial' for Android/Vega entries, 'id' for Chromium) to pass to interaction tools, and to see which targets are already running. Returns { devices, avds } where each device carries a 'platform' discriminator ('ios', 'android', 'chromium', or 'vega'); 'avds' lists Android AVDs bootable via boot-device. A Vega VVD is listed under 'devices' whether running or stopped (state 'running'/'stopped'); start a stopped one with boot-device using its 'vvdImage'. Android entries also carry a 'kind' ('emulator' for a local AVD, 'device' for a physical phone on USB / wireless adb). Physical phones use the same interaction tools as emulators and need no boot-device (just connect the phone with USB debugging authorised). Physical iPhones appear as iOS entries with kind 'device' (no iPads); no boot-device. State 'connected' = cabled and usable; 'paired' = not reachable over USB, never auto-bound. TV targets are tagged with runtimeKind 'tv' (Apple TV simulators, Android TV / leanback devices). They are focus-driven, not touch-driven: use `describe` to read focus, `tv-remote` for remote presses, and `keyboard` to type, rather than the coordinate/gesture tools. iOS simulators from an additional CoreSimulator device set (the 'ios.additionalDeviceSets' configuration, e.g. devices created by Radon IDE) are tagged with their 'deviceSet' path; the same tools drive them by udid, but they run headless (no Simulator.app window). A foldable iOS simulator (the iPhone Duo) or Android emulator (and its AVD) has 'foldable: true'; use the `fold` tool to fold or unfold it. Chromium apps are discovered by probing CDP debugging ports (default 9222; extend via the ARGENT_CHROMIUM_PORTS=<comma-separated-ports> env var). They must already be running with --remote-debugging-port=<port>; u…

Input schema present but exposes no named parameters.

No output schema declared.

No examples provided.

native-find-views ~555

Search for specific UIViews in the running app by class name, accessibility identifier, accessibility label, tag or React Native nativeID; at least one of these params is required. Use when you need to locate a specific view by its properties without dumping the entire native-full-hierarchy. Returns { status: "ok", matches }: the matching views with their frames, properties, optional ancestors and optional children. Frames are in points. frame and center are relative to the superview, and windowFrame turns with the UI. To tap a view, divide the centre of its screenFrame by the screen size in points, which native-full-hierarchy returns as screen (maxDepth: 1 is enough), and pass the result to gesture-tap. Matches are not filtered for visibility. Before tapping, call native-view-at-point with hitTest: true at that centre, with skipClasses ["UITextEffectsWindow"]. If it does not return the match or a descendant, check a screenshot. If status is restart_required: follow the message (usually restart-app), then retry. If status is service_stale: the app is already injected, so restarting it cannot help. Restart the tool-server (`argent server stop && argent server start --detach`) and retry. If the same status comes back after that restart, stop restarting: follow the message, which names the terminal fallback. If status is connect_pending: the app is injected and still connecting. Do not restart it, wait a few seconds and retry. If status is init_failed: follow the message (re-boot the simulator) rather than retrying. An Apple system app is rejected outright (terminal: never retry it).

NameTypeReqDescription
bundleIdstringyesBundle ID of the app
classNamestring–UIView class name to match (exact, e.g. UIButton)
fieldsarray–View fields to include. Defaults: pointer, className, tag, frame, windowFrame, screenFrame, bounds, hidden, alpha, opaque, clipsToBounds, userInteractionEnabled, depth, identifier, label, layerName,…
identifierstring–Accessibility identifier to match (exact)
includeAncestorsboolean–Include ancestor chain for each matched view (default true)
includeChildrenboolean–Include child views for each matched view (default true)
labelstring–Accessibility label to match (exact)
nativeIDstring–React Native nativeID prop to match (exact)
taginteger–UIView tag integer to match
udidstringyesSimulator UDID

No output schema declared.

No examples provided.

native-full-hierarchy ~541

Get the complete UIKit view tree for the running app. WARNING: Output can be extremely large (100KB–500KB+) for complex apps, especially those built with SwiftUI. Prefer native-find-views for targeted queries. Reduce the output with skipClasses / skipClassPrefixes, fields and maxDepth. Use when you need deep layout debugging, finding views with no accessibility labels, or verifying view structure not exposed through the accessibility tree. Returns { status: "ok", windows, screen } with the full view hierarchy. screen gives the screen size in points on its portrait axes, and the interfaceOrientation of the UI (portrait, portraitUpsideDown, landscapeLeft, landscapeRight). Each screenFrame uses the portrait axes of the screen. Each windowFrame uses the axes of its window, which turn with the UI. If status is restart_required: follow the message (usually restart-app), then retry. If status is service_stale: the app is already injected, so restarting it cannot help. Restart the tool-server (`argent server stop && argent server start --detach`) and retry. If the same status comes back after that restart, stop restarting: follow the message, which names the terminal fallback. If status is connect_pending: the app is injected and still connecting. Do not restart it, wait a few seconds and retry. If status is init_failed: follow the message (re-boot the simulator) rather than retrying. An Apple system app is rejected outright (terminal: never retry it).

NameTypeReqDescription
bundleIdstringyesBundle ID of the app
fieldsarray–View fields to include. Use EXACT names. Defaults: pointer, className, tag, frame, windowFrame, screenFrame, bounds, hidden, alpha, opaque, clipsToBounds, userInteractionEnabled, depth, identifier, l…
maxDepthinteger–Maximum recursion depth (default 8). Raise it to reach deeper views. Must be greater than 0.
skipClassPrefixesarray–Class name prefixes to prune entire subtrees. For SwiftUI apps use ["_TtGC7SwiftUI"] to drop mangled SwiftUI generic type subtrees while keeping _UIHostingView and UIKit bridges. Avoid broad prefixes…
skipClassesarray–Exact UIView class names whose entire subtree should be pruned (e.g. ["UIImageView"] to drop image leaf nodes)
udidstringyesSimulator UDID

No output schema declared.

No examples provided.

native-network-logs ~835

Retrieve network requests captured at the native level: NSURLProtocol on iOS, OkHttp on Android. Use when inspecting native-level HTTP traffic that view-network-logs cannot see: requests that never pass through JS XMLHttpRequest or fetch. On iOS it sees URLSession requests (default and ephemeral session configurations, and URLSession.shared) from native modules and Swift/Objective-C code. It does not see background URLSession transfers, WKWebView, Network.framework, or HTTP stacks that bypass URLSession (e.g. Flutter), so an empty result does not mean the app sent nothing. Capture starts at the first call, so the first call is usually empty: repeat the action, then call again. On iOS: returns { status, count, events } where each event contains URL, method, status code, headers, and timing. If status is restart_required: follow the message (usually restart-app), then retry. If status is service_stale: the app is already injected, so restarting it cannot help. Restart the tool-server (`argent server stop && argent server start --detach`) and retry. If the same status comes back after that restart, stop restarting: follow the message, which names the terminal fallback. If status is connect_pending: the app is injected and still connecting. Do not restart it, wait a few seconds and retry. If status is init_failed: follow the message (re-boot the simulator) rather than retrying. On a device whose devtools agent argent attached to rather than armed itself, a not-connected app fails with `NATIVE_DEVTOOLS_NOT_CONNECTED` instead of a status, and no restart on argent's side fixes it. An Apple system app is rejected outright (terminal: never retry it). On Android: needs a debuggable build. The first call starts capture and usually returns armed: false. Call it again every few seconds until armed is true, then repeat the action whose requests you need: earlier requests are not captured. Capture also misses a request already in flight at the attach and requests from clients oth…

NameTypeReqDescription
bundleIdstringyesApp bundle ID on iOS or package name on Android
clearboolean–Clear the log after reading. On Android, the IDs of cleared requests still work in view-network-request-details. Default: false.
limitnumber–Maximum number of entries to return: the most recent ones, oldest first. Default: 50.
portinteger–Android only: the app's Metro port, whose requests are left out of the listing. Omit it to use this device's port, 8081 by default. Range: 1 to 65535.
stopboolean–Android only: turn native capture off for this app instead of listing. Other apps keep theirs.
udidstringyesDevice ID from list-devices: an iOS simulator UDID or Android serial

No output schema declared.

No examples provided.

native-profiler-analyze ~115

Summarize a native trace saved by native-profiler-stop or restored with profiler-load. Returns a Markdown report with severity and suggested next steps. Instruments reports CPU hotspots, UI hangs, and memory leaks on iOS. Android reports also include jank reasons, main-thread state, garbage collection, and memory growth as a weak leak signal. Use profiler-stack-query for details behind a finding.

NameTypeReqDescription
device_idstringyesTarget device id from `list-devices` (iOS UDID or Android serial).

No output schema declared.

No examples provided.

native-profiler-start ~296

Start native profiling. Instruments records CPU, hangs, and leaks on iOS. Perfetto records CPU, jank, and memory growth on Android. Start the target app first for an attach recording. Wait for this call to return before reproducing the scenario, then call native-profiler-stop while the app is still running. On Android, a warning names apps whose timings may be affected by native network capture.

NameTypeReqDescription
app_processstring–iOS: the CFBundleExecutable or display name of the app to profile. Android: the app's package name. If omitted, auto-detects the foreground app; set it only if auto-detection picks the wrong app.
device_idstringyesTarget device id from `list-devices` (iOS UDID or Android serial).
malloc_stack_loggingboolean–iOS only. Enable allocation stacks from app launch. Restarts the app, clears its current state, and adds CPU and memory overhead. Leave it off for CPU and hang timing. On iOS, attach recordings typic…
template_pathstring–iOS only. Path to an Instruments .tracetemplate file, or the name of a built-in template such as "Time Profiler" (defaults to the bundled Argent template).

No output schema declared.

No examples provided.

native-profiler-stop ~92

End a native profiling session started by native-profiler-start and save the trace as a downloadable artifact. On iOS, export CPU, hangs, and leaks XML from Instruments. Check warning and exportDiagnostics for partial or skipped exports. On Android, save the Perfetto .pftrace.

NameTypeReqDescription
device_idstringyesTarget device id from `list-devices` (iOS UDID or Android serial).

No output schema declared.

No examples provided.

native-view-at-point ~709

Inspect the deepest UIView at a raw native window point: the one drawn there, or with hitTest: true the one that would receive a touch. Without hitTest it ignores userInteractionEnabled and answers "what is visually here?". Use when a screenshot shows something the accessibility tree does not name (an unlabeled icon, a decorative overlay, a custom-drawn cell) and you need the class, identifier or nativeID of whatever draws it. With hitTest: true it answers "what would receive the touch?". Use when a tap lands somewhere unexpected or does nothing, to see which control UIKit would hand the touch to: a transparent overlay swallowing it, a parent recognizer, a disabled button. Returns { status: "ok", view }: the matched view with its class name, frames, identifier, label and layer name, plus its ancestors and subviews as requested. view is null when nothing is drawn at that point, or with hitTest when no touchable view sits under it. To get x and y from a normalized describe frame, multiply its values by the screen size in points, also when the UI is landscape. native-full-hierarchy returns that size as screen (pass maxDepth: 1 and fields: ["className"] to keep it small). If status is restart_required: follow the message (usually restart-app), then retry. If status is service_stale: the app is already injected, so restarting it cannot help. Restart the tool-server (`argent server stop && argent server start --detach`) and retry. If the same status comes back after that restart, stop restarting: follow the message, which names the terminal fallback. If status is connect_pending: the app is injected and still connecting. Do not restart it, wait a few seconds and retry. If status is init_failed: follow the message (re-boot the simulator) rather than retrying. An Apple system app is rejected outright (terminal: never retry it).

NameTypeReqDescription
bundleIdstringyesBundle ID of the app
fieldsarray–View fields to include. Defaults: pointer, className, tag, frame, windowFrame, screenFrame, bounds, hidden, alpha, opaque, clipsToBounds, userInteractionEnabled, depth, identifier, label, layerName,…
hitTestboolean–true returns the view that would receive a touch at the point: it respects userInteractionEnabled, close to UIKit hit-testing. Default false: the deepest visible view.
includeAncestorsboolean–Include ancestor chain for the matched view (default true)
includeChildrenboolean–Include child views for the matched view (default false)
maxDepthinteger–Maximum depth for returned child/ancestor serialization (default 150). Must be greater than 0.
skipClassPrefixesarray–Class name prefixes to prune entire subtrees
skipClassesarray–Exact UIView class names whose entire subtree should be pruned
udidstringyesSimulator UDID
xnumberyesRaw X coordinate in points, on the portrait axes of the screen. NOT normalized [0,1] tap space.
ynumberyesRaw Y coordinate in points, on the same axes as x. NOT normalized [0,1] tap space.

No output schema declared.

No examples provided.

open-url ~349

Open a URL or URL scheme on the device. Use when navigating to a web page or deep-linking into an app. On Chromium, this navigates the primary renderer to the given URL; a URL that does not load a page, start a download or run as javascript:, such as tel:, mailto:, an unknown scheme or no scheme, fails there. Cross-platform schemes on native devices: https://, tel:, mailto:. iOS also: messages://, settings://, maps://. Android also: geo:, plus any app-specific deep link. An https:// link opens the native app only if that app is verified for the link's domain (Universal Links / App Links); even then, an iOS simulator may open Safari. To reliably open an installed app, use its custom scheme (scheme://path) or launch-app. On a physical iPhone, the app that receives the URL becomes the app under automation. Returns { opened, url, note? }. `note` explains the routing caveat when a web URL was opened on a native device. Fails if no app is registered to handle the URI (iOS/Android) or the renderer rejects the navigation (Chromium).

NameTypeReqDescription
bundleIdstring–Physical iOS only: the app that receives the URL. Defaults to Safari for http(s); required for any other scheme.
udidstringyesTarget device id from `list-devices` (iOS UDID, Android serial, or Chromium id).
urlstringyesURL or scheme to open (e.g. https://example.com, messages://, tel:555, geo:37.0,-122.0).

No output schema declared.

No examples provided.

paste ~311

Set the DEVICE clipboard to `text`, then trigger the platform's paste shortcut to paste it into the focused field (iOS simulator, Android emulator). Do NOT use this in place of `keyboard`. `keyboard` types as a user would and is the default for all text entry; call this only where a real user would paste: a 2FA/OTP code copied from another app, a long link or token, or when testing the app's own paste handling. Tap the field first so it has focus. Returns { pasted: true }, plus via: "keyboard" when a secret was typed instead of pasted. Fails on a TV target, when the device clipboard cannot be set, when a secret that has to be typed holds a character that cannot be typed, or when the simulator-server build lacks clipboard support.

NameTypeReqDescription
textstringyesText to put on the device clipboard and paste into the focused field. Supports `{{secret:<NAME>}}` placeholders, resolved on the tool-server from the `ARGENT_SECRET_<NAME>` environment variable or an…
udidstringyesTarget device id from `list-devices` (iOS simulator UDID or Android emulator serial).

No output schema declared.

No examples provided.

profiler-combined-report ~175

Generate a cross-correlated report combining React Profiler and native profiler data. Maps native hangs to React commits using wall-clock time alignment. Call this tool when both profilers were run in parallel on the same session. Returns a markdown report correlating hangs with React commits, memory leaks, and investigation hints. Fails if either react-profiler-analyze or native-profiler-analyze has not been called first, or if the session stored no React commits.

NameTypeReqDescription
device_idstringyesDevice ID from list-devices used for the native profile (iOS simulator UDID, Android serial).
portinteger–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.

profiler-commit-query ~382

Query React commit data for iterative investigation of render performance. Modes: - by_component: All commits where component_name rendered, with causes and durations. - by_time_range: What happened in time_range_ms. - by_index: Full detail dump of the commit at commit_index (all components, props changed, parent cascade). - cascade_tree: Parent-child cascade tree for the commit at commit_index, showing who triggered whom. - render_counts: Re-render count, rate and self time per component over ALL commits, including those under 16ms that the other modes cannot see. Library components included. Use when drilling into specific components or time windows after react-profiler-analyze. Returns a markdown table or tree of commit data matching the requested mode. Fails if react-profiler-stop has not been called.

NameTypeReqDescription
commit_indexinteger–Commit index
component_namestring–Component name; optional for render_counts, which it narrows to one component
device_idstringyesDevice ID used to record or load the React profile.
modestringyesQuery mode; see Modes in the tool description.
portinteger–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.
time_range_msobject–Time range in ms since profiling started, the clock this tool prints
top_ninteger–Max results to return. Defaults to 20 for by_component / by_time_range, which count commits, and render_counts, which counts components. by_index counts individual fibers and returns all of them unle…

No output schema declared.

No examples provided.

profiler-cpu-query ~440

Query Hermes CPU profile data with targeted modes for iterative investigation. Modes: - top_functions: Global CPU hotspots ranked by self-time. Optional time_window_ms to filter. - time_window: CPU breakdown for time_window_ms (e.g. during a slow commit or hang). - call_tree: For a given function_name, show its callees, and its callers with include_callers. - component_cpu: For a given component_name, aggregate CPU activity across its stored commits. In top_functions and time_window, self-times are the summed sampling intervals of the samples that landed in the window, so they measure sampled coverage rather than the window's width and do not change if you widen the query. Their tables state how many samples they cover and how much was idle. Each row is one call path, so a function can have several rows and top_n counts rows. Use when investigating JS CPU hotspots or correlating CPU cost with specific components. Returns a markdown table of CPU hotspots, call tree, or per-component CPU breakdown. Fails if no CPU profile is stored; run react-profiler-stop (and ideally react-profiler-analyze) first.

NameTypeReqDescription
component_namestring–Component name
device_idstringyesDevice ID used to record or load the React profile.
function_namestring–Function name
include_callersboolean–Also show callers of the function. Default: false.
modestringyesQuery mode; see Modes in the tool description.
portinteger–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.
react_onlyboolean–top_functions and time_window: list only React component functions (PascalCase names). Default: false.
time_window_msobject–Time window in ms since profiling started, the same clock profiler-commit-query prints
top_ninteger–Number of results to return. Default: 15. Must be greater than 0.

No output schema declared.

No examples provided.

profiler-load ~379

Fetch and restore a profiling session that react-profiler-stop or native-profiler-stop saved to disk, so the query tools (profiler-cpu-query, profiler-commit-query, profiler-stack-query) can read it. Use when you need to revisit past session data without capturing a new recording. Modes: - list: Show all available profiling sessions in the project's debug directory. - load_react: Load a React profiler session (CPU profile + commit tree) into memory. Requires session_id. - load_native: Restore native trace data (xctrace XML on iOS, .pftrace on Android). Requires session_id. Loading replaces the session in memory for this device_id (and Metro port, for React): the one the query tools and react-profiler-analyze read. The replaced session stays on disk; load it again by its session_id. Returns a summary of the loaded session or a session list for the list mode. Fails if the session_id is not found or the session's trace files are missing from disk.

NameTypeReqDescription
app_processstring–Android package name to pass when restoring an older load_native .pftrace session that has no metadata sidecar.
device_idstringyesTarget device id from `list-devices`. Used to cache the loaded React session under the correct port+device key, and required to resolve the native profiler session for load_native.
modestringyesWhat to do; see Modes in the tool description.
portinteger–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.
session_idstring–Timestamp-based session identifier (e.g. '20250313-143022') from the list output.

No output schema declared.

No examples provided.

profiler-stack-query ~267

Inspect a native profile after native-profiler-analyze or profiler-load. Modes: - hang_stacks: Full CPU context during a specific hang (by hang_index). - function_callers: Who calls function_name and what it calls. - thread_breakdown: CPU time split by thread, optionally filtered. - leak_stacks: Memory leak details (iOS only). Requires a trace with leak data.

NameTypeReqDescription
device_idstringyesTarget device id from `list-devices` (iOS simulator UDID or Android serial).
function_namestring–Native function name
hang_indexinteger–0-based index into the hang list
modestringyesQuery mode; see Modes in the tool description.
object_typestring–Object type filter
threadstring–Thread filter. thread_breakdown: case-insensitive substring match. function_callers: exact raw thread name (e.g. ".blueskyweb.app"), or "main" for the UI thread; omit to search ALL threads (each resu…
top_ninteger–Max results to return. Default: 15. Must be greater than 0.

No output schema declared.

No examples provided.

react-profiler-analyze ~320

Analyze stored profiling data and return a markdown performance report. Returns { report, reportFile, hotCommitsTotal, hotCommitsShown, sessionFiles }. The report is structured around hot React commits (≥16ms absolute floor) with per-commit render cascades, root cause identification, and a top components table. Raw profiling data is saved to disk with a unique session timestamp for later reload via profiler-load. After presenting the report, ask the user whether to investigate further (drill-down with profiler-cpu-query / profiler-commit-query) or implement fixes and re-profile for comparison. Use when the profiling session is complete and you need to interpret the collected data. Fails if react-profiler-stop has not been called or no profiling data is stored.

NameTypeReqDescription
annotationsarray–User actions to annotate the report's commits with: each commit shows the action that preceded it.
device_idstringyesDevice ID used to record or load the React profile.
platformstring–Platform for source lookup when the recording does not identify one.
portinteger–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.
project_rootstringyesAbsolute path to the RN project root for session context detection
rn_versionstring–React Native version (e.g. "0.73.4"). Default: "unknown".

No output schema declared.

No examples provided.

Install

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

# add to Claude Code
claude mcp add software-mansion-argent -- npx -y @swmansion/argent mcp
Common questions

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.