Skip to content

MCPServer.md

Latest commit

 

History

History
114 lines (75 loc) · 5.37 KB

MCPServer.md

File metadata and controls

114 lines (75 loc) · 5.37 KB

TRIGGERcmd MCP Server — User Guide

The TRIGGERcmd MCP server lets AI assistants — Claude, ChatGPT, or any tool that speaks the Model Context Protocol — see your TRIGGERcmd commands and run them on your computers, just by chatting in plain English. For example, you can ask your assistant "What can you run on my office computer?" or "Turn off the basement lights," and it will look up the right TRIGGERcmd command and trigger it for you.

It's already running for every TRIGGERcmd account — there's nothing to install on your computer to use it.

Connecting an AI assistant

There are two ways to connect, depending on what your AI tool supports.

Option 1 — Streamable HTTP (recommended)

Most modern AI products that support "remote MCP servers" or "connectors" (Claude.ai, ChatGPT, etc.) can talk to TRIGGERcmd directly over the internet. When your AI tool asks for an MCP server URL, give it:

https://www.triggercmd.com/mcp

The server supports standard OAuth2, so when you add it as a connector your AI tool will open a TRIGGERcmd sign-in/authorize page automatically. Approve it once, and the assistant can then see and run the commands on your TRIGGERcmd account.

ChatGPT users: TRIGGERcmd is also available as a ready-made app — just install it from https://chatgpt.com/apps/triggercmd/asdk_app_694c84d39cb881918c6d181ae69e33fc and sign in with your TRIGGERcmd account. No URL or setup required.

Option 2 — Local stdio server

If your AI tool only supports locally-run ("stdio") MCP servers — for example, some Claude Desktop setups — use the companion project instead:

https://github.com/rvmey/triggercmd-mcp-stdio

It runs as a small process on your own computer, signs in with your TRIGGERcmd account token, and exposes the exact same tools described below.

What the assistant can do once connected

Once connected, the assistant gets access to tools scoped to your TRIGGERcmd account — it can see every computer and command you have, and trigger them on your behalf. There are three kinds of tools:

list_commands

Asks "what commands do I have?" It returns every command across your computers, including:

  • which computer it lives on
  • its name
  • its voice trigger phrase (if it has one)
  • its MCP tool description (if it has one — see "Dynamic command tools" below)

Your assistant typically calls this first when you ask something like "What TRIGGERcmd commands do I have?" or "What can you run on my Office PC?" so it knows what's available before doing anything.

run_command

The general-purpose "run anything" tool. It works for any command on any of your computers, whether or not that command has an MCP tool description set up.

The assistant fills in:

  • computer — the name of the computer that owns the command
  • command — the name of the command to run
  • parameters (optional) — extra text to pass along, for commands that accept parameters

You'll typically see this used when you ask the assistant to run something by name, e.g. "Run 'Good Night' on my Bedroom PC" or "Trigger Roku Play on the Living Room computer."

Dynamic per-command tools

For any command where you've filled in an MCP Tool Description, TRIGGERcmd automatically creates a dedicated tool just for that command, named something like run_<computer>_<command> (lower-cased, with spaces and special characters turned into underscores).

For example, a command named "Play Music" on a computer named "Office PC" shows up to the assistant as a tool named run_office_pc_play_music, described using whatever text you put in its MCP Tool Description — for instance, "Plays my work playlist through the office speakers."

Why this is useful: giving a command its own tool with a clear, specific description helps the assistant:

  • find and pick the right command quickly, instead of guessing from a short or generic command name
  • how to format any command parameters, and when to use them
  • explain to you, in plain language, what it's about to do before running it

If the command also has Allow Params turned on, its dynamic tool accepts an optional parameters value, just like run_command does. If Allow Params is off, the dynamic tool takes no input — it simply runs the command.

Giving one of your commands its own tool

To make a command show up to AI assistants as its own named tool:

  1. Open the command's settings (in your local TRIGGERcmd agent's commands.json, or wherever you manage that command).
  2. Fill in MCP Tool Description with a short, clear sentence describing what the command does and/or when it should be used — e.g. "Turns off all the lights downstairs" or "Restarts the Plex server."
  3. (Optional) Turn on Allow Params if the command should accept text as a command parameter from the assistant.
  4. Save and let the agent sync. Your TRIGGERcmd Commands list will show a 🔧 icon next to any command that has an MCP tool description.

The next time an assistant lists your tools, it will see a dedicated tool for that command, described the way you wrote it.

Tips for writing good descriptions

  • Say what happens: "Turns on the basement lights, or adjusts brightness. The parameter can be a brightness percentage like 50," not just "Lights."
  • Keep it short — one or two sentences is plenty. The assistant reads it every time it decides whether to use the tool.