Skip to main content
Glama

Unified application task

app_task
Destructive

Use this when the user needs to execute one typed intent inside a curated app via trusted local API, scoped Accessibility or CDP without screenshots. New local- IDs support inspect/find-visible-controls and explicitly scoped non-destructive focus_control, after unique PID/window and semantic ID/name/role checks. Unknown apps cannot launch, click, invoke or submit; a single native text value can change only after explicit commit and exact old-value check (raw expected_text or an expiring blind token from inspect/find). Telegram edits require authenticated owner, original text and verified read-back.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "app": "<app>",
      "intent": "<intent>",
      "device": "<device>"
    },
    "name": "app_task"
  }
}

Input Schema

NameRequiredDescriptionDefault
appYesRegistered app id or local-<hash> inventory id for read-only inspect/find of visible controls; unknown apps support inspect, visible lookup, verified focus or explicit one-field text replacement with old-value check; no launch, click or submission.
pidNoExisting positive native process id to disambiguate the window.
queryNoWith registered-app find: text entered into a verified search field. With detected local- app find: read-only search of already visible accessibility names, no typing. With find_post: read-only Telegram query through an authorized local API; never selects or edits a result automatically.
commitNoWith edit_post or set_control_text: explicit authorization to change exactly the verified post or native text value; requires expected_text and replacement_text.
deviceYesReMCP device id returned by list_devices. Required for every device operation.
intentYesHigh-level user intent, no untrusted selectors, code or shortcuts.
verifyNoFor native app launch only, inspect a real app window after dispatch; may take longer.
channelNoWith find_post only: optional exact public Telegram channel username, e.g. @mychannel. Never a substring or guessed channel.
dry_runNoReturn semantic route/transaction plan without opening apps, typing or editing; redact user-provided text.
post_urlNoWith edit_post: exact https://t.me/channel/messageId permalink; never a guessed post.
target_idNoExisting exact Chromium tab id; useful when multiple authorized tabs match.
window_idNoExisting exact native window id to disambiguate the window.
control_idNoFor focus_control/set_control_text: exact accessibility element ID from the verified native inspect/find result.
expected_nameNoFor focus_control/set_control_text: exact previously observed accessible name.
expected_roleNoFor focus_control/set_control_text: exact previously observed accessible role. No invoke/submit.
expected_textNoFor edit_post: exact previously read post text. For set_control_text: exact old native field text, or supply expected_value_token instead.
replacement_textNoWith edit_post: bounded post replacement, max 4096. With set_control_text: 1–256 single-line non-secret characters; never presses Enter.
expected_value_tokenNoFor set_control_text: blinded 90-second field-version token from a verified app_task inspect/find response; alternative to revealing expected_text. Bound to app/PID/window/control.

Output Schema

NameRequiredDescriptionDefault
dataNoArray/scalar payload wrapper used when the natural result is not an object.
textYesConcise human-readable result of the tool call.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Added

TDQS

B3.3/5.0
Behavior4/5

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

Annotations already declare destructive=true, openWorld=true and non-idempotent, and the description adds meaningful guardrails: no screenshots, no launch/click/invoke/submit on unknown apps, mutation only after explicit commit plus exact old-value check, and Telegram edits requiring an authenticated owner with verified read-back. The 90-second blind-token lifetime and PID/window/role checks are useful risk context. It stops short of describing failure modes or rate/authorization behavior in general.

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

Conciseness3/5

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

The use condition is front-loaded, which is good, but the body is a single dense block of jargon ('New local- IDs support inspect/find-visible-controls', 'expiring blind token') with no structure for 18 parameters and 8 intents. Size is defensible for a tool this complex, but readability suffers and key constraints blend together.

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?

For a destructive, open-world tool with 18 parameters, an output schema, and rich annotations, the description covers the critical behaviors: scope limits, mutation prerequisites, verification and token expiry. Gaps remain in intent-by-intent behavior (e.g. what dry_run returns, how open vs inspect differ), but the core safety envelope is present.

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 100%, so the schema already documents all 18 parameters and the baseline of 3 applies. The description adds real cross-parameter meaning — the dependency chain of commit + expected_text (or expiring token) + replacement_text for text changes, and the local- ID path to inspect/find — but most per-parameter semantics remain in the schema rather than the description.

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

Purpose3/5

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

The description names a verb and resource ('execute one typed intent inside a curated app') and a mechanism (trusted local API, scoped Accessibility or CDP, no screenshots), which is more concrete than a tautology. But 'typed intent' is jargon that only resolves after reading the schema's intent enum, and nothing distinguishes it from siblings like app_execute, app_navigate or launch_app. Purpose is inferable but requires cross-referencing.

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?

It gives a positive usage condition ('Use this when the user needs to execute one typed intent inside a curated app') and states hard exclusions for unknown apps ('cannot launch, click, invoke or submit'). It does not route between this tool and the many sibling application/browser tools, so the agent must guess which entry point to pick.

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

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources