llm-mom

exrhizo/llm-mom

3.1

If you are the rightful owner of llm-mom and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.

LLM Mom is an MCP accountability agent designed to manage long-running workflows in tmux panes by injecting commands when the pane is idle.

LLM Mom

An MCP accountability agent that keeps LLMs working until they actually finish the job instead of stopping prematurely.

Setup

Requirements: Python 3.12+, tmux, OpenAI API key

# Install as a tool (works from any directory)
uv tool install --editable . --force

# Ensure ~/.local/bin is on your PATH
export PATH="$HOME/.local/bin:$PATH"

# Configure (create .env in the repo)
OPENAI_API_KEY=sk-...
MODEL=openai:gpt-4.1-nano

Usage

Add to Claude Code or Claude Desktop:

claude mcp add --transport stdio mom \
  --env OPENAI_API_KEY="$OPENAI_API_KEY" \
  --env MODEL="${MODEL:-openai:gpt-4.1-nano}" \
  -- llm-mom stdio

In Claude, start the watcher with hey_mom:

Use mom.hey_mom to start accountability on my pane.
Goal: run tests and fix any failures until they pass
Pane: %1 (from tmux display-message -p '#{pane_id}')
wait_cmd: "tail -n 50 logs/app.log"

Mom runs a background loop that waits for your pane to be idle, asks an LLM if the goal is done, and injects commands to continue. Use mom.look_at_this to add context while it's running. Use mom.reset to stop.