agent-device

Official

Overview Schema Source Related Servers Discussions Admin

perf

Collects performance profiling artifacts and generates compact summaries for CPU, memory, and frame data to support performance analysis.

Instructions

Collect frame health, memory diagnostics, and platform profiling artifacts with compact agent-readable summaries. For CPU profiles, start and stop write the raw artifact while report writes a compact summary; request the report when the task needs readable native CPU evidence. Profiling output is evidence only: compact state, artifact path, and size. Times out after 90s.

Input Schema

TableJSON Schema

Name	Required	Description
`out`	No	Output artifact path.
`area`	Yes
`kind`	No
`udid`	No	Apple device or simulator UDID; the selector that pins one device when several share a name.
`debug`	No	Enable debug diagnostics.
`runId`	No	Lease run identifier.
`action`	No
`device`	No	Device name selector (a UDID belongs in udid, a serial in serial).
`serial`	No	Android, HarmonyOS, or Vega VVD serial selector.
`target`	No	Alias for deviceTarget on commands without a UI target field. Interaction commands reserve target for the UI element.
`tenant`	No	Remote tenant identifier.
`leaseId`	No	Existing lease identifier.
`session`	No	Agent-device session name.
`subject`	No
`noRecord`	No	Do not record this action.
`platform`	No	Platform selector used to resolve a device.
`template`	No	xctrace template name, for example Time Profiler.
`tracePath`	No	Existing .trace path to report, defaults to the latest session trace.
`includeCost`	No	Include per-command agent-cost (cost.wallClockMs, …) in structuredContent. Defaults to off; the default response shape is unchanged.
`deviceTarget`	No	Device target form. Maps to the CLI --target flag.
`responseLevel`	No	Response verbosity: token-cheap digest / default (today) / full. Defaults to default; the default response shape is unchanged.
`mcpOutputFormat`	No	MCP text content format. Defaults to optimized agent-friendly text; use json for JSON text. Structured content is always returned separately.
`androidDeviceAllowlist`	No	Android serial allowlist used for device resolution.

Schema Changelog

Changes observed during successful MCP inspections.

Changed3 schema fields changedv0.21.0Sep 9, 2026

changedInput schema / properties / device / description

Previous value: -"Device name selector."New value: +"Device name selector (a UDID belongs in udid, a serial in serial)."

changedInput schema / properties / serial / description

Previous value: -"Android device or Vega VVD serial selector."New value: +"Android, HarmonyOS, or Vega VVD serial selector."

changedInput schema / properties / udid / description

Previous value: -"iOS device UDID selector."New value: +"Apple device or simulator UDID; the selector that pins one device when several share a name."

Changed10 schema fields changedv0.20.10Aug 26, 2026

changedInput schema / properties / area / enum

Previous value: -[
-  "metrics",
-  "frames",
-  "memory",
-  "cpu",
-  "trace"
-]New value: +[
+  "frames",
+  "memory",
+  "cpu",
+  "trace"
+]

removedInput schema / properties / cwd

Removed value: -{
-  "description": "Working directory for command execution.",
-  "type": "string"
-}

removedInput schema / properties / daemonAuthToken

Removed value: -{
-  "description": "Remote daemon auth token.",
-  "type": "string"
-}

removedInput schema / properties / daemonBaseUrl

Removed value: -{
-  "description": "Remote daemon base URL.",
-  "type": "string"
-}

removedInput schema / properties / iosSimulatorDeviceSet

Removed value: -{
-  "description": "iOS simulator device-set path used for device resolution.",
-  "type": "string"
-}

removedInput schema / properties / iosXctestDerivedDataPath

Removed value: -{
-  "description": "Derived data path for external iOS XCTest runner execution.",
-  "type": "string"
-}

removedInput schema / properties / iosXctestEnvDir

Removed value: -{
-  "description": "Writable directory for iOS XCTest runner env overlays.",
-  "type": "string"
-}

removedInput schema / properties / iosXctestrunFile

Removed value: -{
-  "description": "Externally built iOS XCTest runner .xctestrun artifact path.",
-  "type": "string"
-}

removedInput schema / properties / stateDir

Removed value: -{
-  "description": "Agent-device state directory.",
-  "type": "string"
-}

addedInput schema / required
```
Added value: +[
+  "area"
+]
```

Changed1 schema field changedv0.20.9Aug 19, 2026

changedInput schema / properties / platform / enum

Previous value: -[
-  "apple",
-  "android",
-  "vega",
-  "linux",
-  "web",
-  "ios",
-  "macos"
-]New value: +[
+  "apple",
+  "android",
+  "harmonyos",
+  "vega",
+  "linux",
+  "web",
+  "ios",
+  "macos"
+]

Changed3 schema fields changedv0.20.2Jul 30, 2026

addedInput schema / properties / noRecord

Added value: +{
+  "description": "Do not record this action.",
+  "type": "boolean"
+}

changedInput schema / properties / platform / enum

Previous value: -[
-  "apple",
-  "android",
-  "linux",
-  "web",
-  "ios",
-  "macos"
-]New value: +[
+  "apple",
+  "android",
+  "vega",
+  "linux",
+  "web",
+  "ios",
+  "macos"
+]

changedInput schema / properties / serial / description

Previous value: -"Android serial selector."New value: +"Android device or Vega VVD serial selector."

Changed3 schema fields changedv0.17.11Jul 6, 2026

addedInput schema / properties / includeCost

Added value: +{
+  "description": "Include per-command agent-cost (cost.wallClockMs, …) in structuredContent. Defaults to off; the default response shape is unchanged.",
+  "type": "boolean"
+}

changedInput schema / properties / platform / enum

Previous value: -[
-  "ios",
-  "macos",
-  "android",
-  "linux",
-  "web",
-  "apple"
-]New value: +[
+  "apple",
+  "android",
+  "linux",
+  "web",
+  "ios",
+  "macos"
+]

addedInput schema / properties / responseLevel

Added value: +{
+  "description": "Response verbosity: token-cheap digest / default (today) / full. Defaults to default; the default response shape is unchanged.",
+  "enum": [
+    "digest",
+    "default",
+    "full"
+  ],
+  "type": "string"
+}

Changed1 schema field changedv0.17.7Jun 23, 2026

changedInput schema / properties / platform / enum

Previous value: -[
-  "ios",
-  "macos",
-  "android",
-  "linux",
-  "apple"
-]New value: +[
+  "ios",
+  "macos",
+  "android",
+  "linux",
+  "web",
+  "apple"
+]

Changed12 schema fields changedv0.17.6Jun 18, 2026

changedInput schema / properties / action / enum

Previous value: -[
-  "sample"
-]New value: +[
+  "sample",
+  "snapshot",
+  "start",
+  "stop",
+  "report"
+]

changedInput schema / properties / area / enum

Previous value: -[
-  "metrics",
-  "frames"
-]New value: +[
+  "metrics",
+  "frames",
+  "memory",
+  "cpu",
+  "trace"
+]

addedInput schema / properties / iosXctestDerivedDataPath

Added value: +{
+  "description": "Derived data path for external iOS XCTest runner execution.",
+  "type": "string"
+}

addedInput schema / properties / iosXctestEnvDir

Added value: +{
+  "description": "Writable directory for iOS XCTest runner env overlays.",
+  "type": "string"
+}

addedInput schema / properties / iosXctestrunFile

Added value: +{
+  "description": "Externally built iOS XCTest runner .xctestrun artifact path.",
+  "type": "string"
+}

addedInput schema / properties / kind

Added value: +{
+  "enum": [
+    "xctrace",
+    "simpleperf",
+    "perfetto",
+    "android-hprof",
+    "memgraph"
+  ],
+  "type": "string"
+}

addedInput schema / properties / mcpOutputFormat

Added value: +{
+  "description": "MCP text content format. Defaults to optimized agent-friendly text; use json for JSON text. Structured content is always returned separately.",
+  "enum": [
+    "optimized",
+    "json"
+  ],
+  "type": "string"
+}

addedInput schema / properties / out

Added value: +{
+  "description": "Output artifact path.",
+  "type": "string"
+}

changedInput schema / properties / platform / enum

Previous value: -[
-  "ios",
-  "android",
-  "macos",
-  "linux",
-  "apple"
-]New value: +[
+  "ios",
+  "macos",
+  "android",
+  "linux",
+  "apple"
+]

addedInput schema / properties / subject

Added value: +{
+  "enum": [
+    "profile"
+  ],
+  "type": "string"
+}

addedInput schema / properties / template

Added value: +{
+  "description": "xctrace template name, for example Time Profiler.",
+  "type": "string"
+}

addedInput schema / properties / tracePath

Added value: +{
+  "description": "Existing .trace path to report, defaults to the latest session trace.",
+  "type": "string"
+}

Changed2 schema fields changedv0.17.1Jun 9, 2026

addedInput schema / properties / action

Added value: +{
+  "enum": [
+    "sample"
+  ],
+  "type": "string"
+}

addedInput schema / properties / area

Added value: +{
+  "enum": [
+    "metrics",
+    "frames"
+  ],
+  "type": "string"
+}

First observedv0.16.9Jun 3, 2026

TDQS

A3.8/5.0

Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations provided, the description carries the full burden. It discloses the 90-second timeout, describes the output as 'evidence only' with 'compact state, artifact path, and size', and clarifies the behavior of start/stop vs report for CPU profiles. This goes beyond a minimal description, though it does not address side effects, permissions, or error conditions.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is three sentences, front-loaded with the primary purpose, then adding action-specific guidance and a timeout note. Every sentence contributes distinct information with no waste. It is compact and well-structured for an agent to parse quickly.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool's complexity (23 parameters) and no output schema, the description covers the essential behavior: what it collects, the distinction between raw and summary artifacts, output shape, and the timeout. It relies on the schema for parameter details, which is reasonable given 83% coverage. It does not mention error handling or edge cases, but the information provided is sufficient for an agent to call it correctly in common cases.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 83%, so the schema already documents most parameters. The description adds value by explaining the action semantics (e.g., start/stop vs report behavior), which informs how parameters like 'action' and 'kind' are used. However, it does not elaborate on the many selector parameters (udid, serial, device), relying on the schema's existing descriptions. This is adequate given high coverage.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool's purpose: 'Collect frame health, memory diagnostics, and platform profiling artifacts with compact agent-readable summaries.' It specifies the resource (profiling artifacts) and the action (collect). It distinguishes actions within the tool (start/stop vs report) but does not explicitly contrast with sibling tools like 'trace', so it stops short of full differentiation.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description provides some guidance on when to use the 'report' action ('request the report when the task needs readable native CPU evidence') and notes that start/stop write raw artifacts. However, it does not compare this tool to alternatives or state when to prefer 'perf' over a sibling like 'trace' or 'artifacts'. The guidance is limited to internal action selection, not tool selection.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Deploy Server

perf

Instructions

Input Schema

Schema Changelog

TDQS

Other Tools