set_always_allowed

Diagnostics & system · macOS · Windows

Effect: write (strongest supported action). Status: gated.

Gated means the tool requires an installed app, a connected account, permission, or runtime availability. Check those requirements on the listed platform before using it.

macOS

Purpose

Turns «allowed without asking» on or off for one kind of action that LMCP otherwise confirms first (sending emails or messages, permanent deletes, terminal commands…). Call it only when the person asked for it in this conversation: allowed: true when they say "don't ask me again" for that kind of action, allowed: false when they want to be asked again. Turning one on needs confirm: true after the person approves; without it the answer says what would change and nothing changes. Turning one off acts at once. lmcp_state lists what is allowed now (always_allowed). Your organization's IT policy can forbid turning one on.

Required inputs

action — The kind of action: send_email (send_email, reply_email, Microsoft 365, Proton, Thunderbird), send_imessage, send_teams, reply_whatsapp, reply_telegram, send_zalo, send_servicenow, send_report, publish_recipe, submit_web_form, run_web_script, run_command, run_shortcut, and the permanent deletes.
allowed — true: do it without asking from now on. false: ask again.

Permissions and confirmation

Explicit confirmation parameter: inspect the schema and tool description before any real call. The documentation example keeps confirm false. Obtain user authorization before performing the action.

Use only the apps, accounts and resources authorized by the user on this platform. Consult this tool's platform-specific description for its connection and permission requirements; this catalog does not assert additional OAuth scopes or OS entitlements.

Full input schema — macOS

{
  "properties": {
    "action": {
      "description": "The kind of action: send_email (send_email, reply_email, Microsoft 365, Proton, Thunderbird), send_imessage, send_teams, reply_whatsapp, reply_telegram, send_zalo, send_servicenow, send_report, publish_recipe, submit_web_form, run_web_script, run_command, run_shortcut, and the permanent deletes.",
      "enum": [
        "delete_event_permanent",
        "delete_file_permanent",
        "delete_message_permanent",
        "delete_recipe_permanent",
        "delete_reminder_permanent",
        "delete_web_session_permanent",
        "publish_recipe",
        "reply_telegram",
        "reply_whatsapp",
        "run_command",
        "run_shortcut",
        "run_web_script",
        "send_agent",
        "send_email",
        "send_imessage",
        "send_report",
        "send_servicenow",
        "send_teams",
        "send_zalo",
        "submit_web_form"
      ],
      "type": "string"
    },
    "allowed": {
      "description": "true: do it without asking from now on. false: ask again.",
      "type": "boolean"
    },
    "confirm": {
      "description": "true after the person approved turning it on. Not needed to turn one off.",
      "type": "boolean"
    }
  },
  "required": [
    "action",
    "allowed"
  ],
  "type": "object"
}

Documentation example

Do not execute this example. These concrete inputs refer to a fictional demonstration dataset. Resolve real handles and obtain user authorization before any real call.

{
  "action": "send_email",
  "allowed": false,
  "confirm": false
}

Windows

Purpose

Turns «don't ask me again» on or off for one action family: with allowed true, LMCP does that action (for example send_email: send emails) without the preview-and-confirm step, from any AI app on this computer; with allowed false it asks again. Turning one on is the person's decision: the call without confirm=true changes nothing and returns what would be allowed; ask the person, then call again with confirm=true. Turning one off acts on the first call. The organisation's policy can forbid it. The person sees and revokes them in LMCP's icon → More actions → Allowed without asking.

Required inputs

action — The action family, as the preview of a tool that asks names it.
allowed — true: do it without asking; false: ask again.

Permissions and confirmation

Explicit confirmation parameter: inspect the schema and tool description before any real call. The documentation example keeps confirm false. Obtain user authorization before performing the action.

Use only the apps, accounts and resources authorized by the user on this platform. Consult this tool's platform-specific description for its connection and permission requirements; this catalog does not assert additional OAuth scopes or OS entitlements.

Full input schema — Windows

{
  "properties": {
    "action": {
      "description": "The action family, as the preview of a tool that asks names it.",
      "enum": [
        "send_email",
        "send_teams",
        "reply_whatsapp",
        "reply_telegram",
        "send_servicenow",
        "send_report",
        "send_agent",
        "submit_web_form",
        "run_web_script",
        "run_command",
        "delete_message_permanent",
        "delete_file_permanent",
        "delete_recipe_permanent",
        "delete_web_session_permanent"
      ],
      "type": "string"
    },
    "allowed": {
      "description": "true: do it without asking; false: ask again.",
      "type": "boolean"
    },
    "confirm": {
      "description": "Required (true) to turn a family ON, after the person agreed. Without it, returns a preview. Not needed to turn one off.",
      "type": "boolean"
    }
  },
  "required": [
    "action",
    "allowed"
  ],
  "type": "object"
}

Documentation example

Do not execute this example. These concrete inputs refer to a fictional demonstration dataset. Resolve real handles and obtain user authorization before any real call.

{
  "action": "send_email",
  "allowed": false,
  "confirm": false
}