Posts made by Russ

Thanks @Frederick1337 !

@wandersonjvs, thanks for the screenshots. I suspect you have some antivirus software that's preventing the agent from creating the .TRIGGERcmdData folder under your user's home folder.

You could try creating it manually, or you could try temporarily disabling your antivirus software.

Add an entry like this under mcpServers in claude_desktop_config.json, or mcp.json for VS Code:

        "triggercmd": {
            "command": "docker",
            "args": [
                "run",
                "-i",
                "--rm",
                "-e",
                "TRIGGERCMD_TOKEN",
                "rvmey/triggercmd-mcp"
            ],
            "env": {
                "TRIGGERCMD_TOKEN": "your triggercmd token"
            }
        },

This is the docker image: https://hub.docker.com/r/rvmey/triggercmd-mcp
This is the github repo: https://github.com/rvmey/triggercmd-mcp-stdio

I used this bash script to run an MCP/OpenAI proxy server in Ubuntu under WSL:

#!/bin/bash -xv
curl -O https://agents.triggercmd.com/triggercmd-mcp/triggercmd-mcp-linux-amd64
chmod +x triggercmd-mcp-linux-amd64

docker build -t mcp-proxy-server .
docker run -it -p 8000:8000 -e TRIGGERCMD_TOKEN="my triggercmd token" mcp-proxy-server

This is my Dockerfile:

FROM python:3.11-slim
WORKDIR /app
RUN pip install mcpo uv

COPY triggercmd-mcp-linux-amd64 /triggercmd-mcp-linux-amd64

# Replace with your MCP server command; example: uvx mcp-server-time
CMD ["uvx", "mcpo", "--host", "0.0.0.0", "--port", "8000", "--api-key", "top-secret", "--", "/triggercmd-mcp-linux-amd64" ]

This is how you set it up in the Settings - External Tools - Manage Tool Servers:

There are 3 ways to use it:

• The chat interface built into the web UI. This is mostly for quickly testing the TRIGGERcmd MCP server, but I think people will find it more useful in a Claude Desktop or some tool like that because it can be used in combination with other MCP servers in that context.
• The local stdio based MCP server for local tools like Claude Desktop or VS Code.
• The Streamable HTTP MCP Server for online LLM AI tools, like Mistral AI.

These are the LLM prompts for some of my favorite use-cases:

• "Use #triggercmd to run calculator on NewLaptop" <- this is a good "hello world" type example.
• "Play a popular Youtube video on my upstairs Roku" <- this one is fun and practical.

You can use this TRIGGERcmd MCP server with online tools like Mistral AI.

To set it up in Mistral AI, go here: https://chat.mistral.ai/connections

Then click + Add Connector and fill in this form:

It works just like the local stdio MCP server used with local AI tools like Claude desktop. Click here for an example of that.

You can also use it with Open AI's agent builder tool at https://platform.openai.com/agent-builder

I'm having fun with the TRIGGERcmd MCP server and this Youtube MCP server. This is an example to demostrate how easy it is to turn any script into an MCP tool for AI to use on your behalf.

Here's a Youtube video showing how it works: https://youtu.be/ctV8TUvcQuI

This shows me telling the Claude AI LLM to lookup a Youtube video and play it on my Roku.

This is my claude_desktop_config.json file:

{
  "mcpServers": {
    "triggercmd": {
      "command": "C:\\tools\\triggercmd-mcp-windows-amd64.exe"
    },
    "youtube": {
            "command": "C:\\Users\\russe\\AppData\\Local\\Microsoft\\WindowsApps\\python.exe",
            "args": [
                "-m", "uv", "run", "--directory", 
                "C:\\tools\\mcp\\youtube-mcp", 
                "youtube_server.py"
            ],
            "env": {
                "YOUTUBE_API_KEY": "(get your own)"
            }
    }
  }
}

This is my TRIGGERcmd command from commands.json:

 {
  "trigger": "Youtube Upstairs Roku",
  "command": "c:\\tools\\yt-upstairs-roku.bat",
  "offCommand": "",
  "ground": "foreground",
  "voice": "",
  "voiceReply": "",
  "allowParams": "true",
  "mcpToolDescription": "Shows a specific video on my upstairs Roku.  The parameter specifies the video by its Youtube video ID. "
 }

This is my yt-upstairs-roku.bat file:

curl -v -XPOST "http://192.168.1.12:8060/launch/837?contentId=%1"

192.168.86.127 is the IP of my upstairs Roku.
837 is the ID of the Youtube app on Roku.
%1 is a placeholder for the parameter (the Youtube video ID).

Here I'm using the triggercmd MCP server in goose:

Now you don't have to install the TRIGGERcmd MCP server in Claude or VSCode if you just want to try it out. You can use it right in your browser, if you're subscribed.

Also, as of agent version 1.0.50, a new mcp tool will be added for any command that has an MCP Tool Description:

Your description should explain to the AI LLM what the command does and how to use the parameters. For example:

"This is a color LED light bulb that you can turn on or off. You can also change the color. The parameters are on, off, or a color like red, blue, green etc."

@Fabio, are you able to run your script as a TRIGGERcmd command yet?

You can run it with a command like this:

python3 c:\folder\pyautogui-position\script.py

This should work with any MCP client, but I've tested it with VS Code and Claude Desktop.

This video shows how to set it up and use it: https://youtu.be/4WSN52Optjk

This screenshot shows the basic functionality, but click here for a more interesting use case that uses a "dynamic tool" that gets created for each command that has an MCP Tool Description.

I put the instructions and download links here:
https://github.com/rvmey/triggercmd-mcp-stdio

Or for an online LLM like Mistral or OpenAI agent builder, use the online TRIGGERcmd MCP server.

@garym77, does that command work from a Terminal window? You should run it there first so you can see if the command gives an error.

Also, you should make the voice word something like creality so you can say "Alexa, turn on creality" after you enable the Triggercmd Smart Home skill.

@Stanisław-Grzybołowski

I emailed you the PayPal transaction info. Hopefully that will serve as an invoice.

For now you can only use it if you run your own n8n community edition server because n8n has not verified it yet. I submitted it to be verified and I'll update this post when/if they do verify it.

@Furkan-Cevik, you're welcome. I just published version 1.0.48. I hope that fixes it, but please tell me if you see it again after upgrading.

@Furkan-Cevik, thank you for reporting it. I have a theory about what's causing it, and I'll produce a new agent version soon.

@Convaiyer, let's talk about it. I just sent you a PM.

@Ron-Hoskins, I upgraded your account to subscribed for now, just so we can troubleshoot this.

The Runs list does not get updated when you're not subscribed and try to run more than one command per minute. Now that your account is subscribed, you should be able to see what command is running more than once per minute.

Let me know what you see please.

@rsxyo, you could try this old mac version. I don't have Yosemite to test it myself, but it might be old enough.

https://triggercmdagents.s3.us-east-1.amazonaws.com/TRIGGERcmdAgent36.dmg

@Strahan, I wonder if when the TC agent runs your batch file, the sql.exe file isn't in your PATH, so it doesn't run. The agent is probably running your batch file, which you've basically already proven with your loghealth.bat experiment.

I like that you added another line to your laundryreset.bat file that writes something to a log file, but it's writing your log.txt file to the present working directory, which in the context of the TC agent is something like this:
C:\Users\russ\AppData\Local\triggercmdagent\app-1.0.47\resources\app\src

Assuming it is running your laundryreset script, you can probably fix this if you put the full path to your sql.exe file (and your log file) in the laundryreset.bat file, something like this:

@echo off
c:\full\path\sql.exe homecenter "delete from activeitems where action = 8192" /u service.triggercmd.homecenter /p (pwd) /h lioth.mydomain.com /l c:\logs\laundryreset.log
echo From laundryreset.bat >> c:\logs\laundryreset.log
dir >> c:\logs\laundryreset.log

You'll want to remove the last 2 lines, especially the dir line, but I put it there because that can show you what the working directory is.