act
Execute a natural language action. The AI will plan and perform multi-step operations in a single invocation, useful for transient UI interactions (e.g., Spotlight, dropdown menus) that disappear between separate commands.
This record as markdown: /tools/android/act.md
What act does on Android
AI agents invoke act to trigger actions in Android. What it does depends on the arguments the agent supplies, and its effects often reach beyond the immediate call: builds kicked off, notifications sent, workflows started.
| Parameter | Type | Required | Description |
|---|---|---|---|
prompt | string | Yes | Natural language description of the action to perform, e.g. "press Command+Space, type Safari, press Enter" |
deepThink | boolean | — | Plan this action with deep thinking (richer context and sub-goal decomposition). Helps with complex multi-step instructions at the cost of speed. Defaults to th |
deepLocate | boolean | — | Use deep locate for every element this action targets. Improves precision for small or ambiguous targets at the cost of speed. Defaults to the server --deep-loc |
android.deviceId | string | — | Android device ID (from adb devices) |
android.useScrcpy | boolean | — | Enable scrcpy accelerated screenshots |
android.aiActContext | string | — | Background knowledge passed to aiAct. Default: no extra context. |
android.waitAfterAction | number | — | Wait time in milliseconds after each action execution. Default: 300ms. |
android.replanningCycleLimit | integer | — | Maximum number of replanning cycles for aiAct. Default: model adapter default. |
android.screenshotShrinkFactor | number | — | Screenshot shrink factor before sending images to AI. Default: 1; high values may reduce recognition quality, especially on mobile. |
Parameters from the server's own tool schema.
Why act is rated High
This tool triggers execution of multi-step automated operations on an Android device based on natural language prompts. While not destructive by itself, it can perform any sequence of device interactions (opening apps, modifying settings, sending messages, etc.) whose actual effects depend on how the natural language is interpreted.
From the tool's definition Tool description states it 'Execute[s] a natural language action' and 'perform[s] multi-step operations', enabling the AI to control Android device interactions autonomously.
Attacks that exploit this kind of access
The rule that runs act safely
PolicyLayer is an MCP gateway: it sits between your AI agents and Android, and checks every tool call against a rule you set before the call runs. Nothing changes on the server itself. For act, this is the rule to start with:
act stays usable, but rate-capped: a runaway agent can't fire it dozens of times a minute. Everything else on the server is denied unless you say otherwise.
The button opens the PolicyLayer dashboard: create your workspace, connect Android, apply this rule, and every act call is checked against it from then on.
Questions about act
Execute a natural language action. The AI will plan and perform multi-step operations in a single invocation, useful for transient UI interactions (e.g., Spotlight, dropdown menus) that disappear between separate commands. It is categorised as a Execute tool in the Android MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
act accepts 9 parameters: prompt, deepThink, deepLocate, android.deviceId, android.useScrcpy, android.aiActContext, android.waitAfterAction, android.replanningCycleLimit, android.screenshotShrinkFactor. Required: prompt. The full parameter table on this page comes from the server's own tool schema.
Register the Android MCP server in PolicyLayer and add a rule for act: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Android. Nothing to install.
act is a Execute tool with high risk. Execute tools should be rate-limited and have argument validation enabled.
Yes. Add a rate_limit block to the act rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for act. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
act is provided by the Android MCP server (@midscene/android-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
More on Android, and thousands of servers like it.
Across the catalogue