☰
Grok BotAutomation & Routines2026-10-14

Let Grok Bot answer your iMessages through an open-source bridge with an allowlist and a kill switch

Demo GIF thumbnail of the iMessage bridge from @tvykruta's post

Tomas Vykruta built imessage-agent-bridge over a weekend and released it under the MIT license. A helper on their Mac checks for new iMessages (every 15 seconds by default) and passes them to their Grok Bot agents, which text people back on their own. Per the README, the Mac enforces a recipient allowlist, 30 texts an hour, a 1,000-character cap and a kill switch, without giving the bots system access to the Mac.

Try it yourself

What you get

A Mac helper that relays new iMessages to your agent and only lets it text approved numbers, within hourly and length limits.

Tools

  • Grok Bot
  • imessage-agent-bridge (MIT)
  • macOS Messages
  • ntfy (default relay)

Time / cost: Not stated; the author says it was built over a weekend

Steps

  1. Install the bridge on your Mac using the README.
  2. Add a short allowlist of numbers your agent may text.
  3. Keep the default limits: 30 texts an hour and 1,000 characters.
  4. Know the kill switch: touch ~/.smsbridge/DISABLED or set "enabled": false.

Copy-ready prompt

Help me install imessage-agent-bridge from https://github.com/tvykruta/imessage-agent-bridge on my Mac. Walk me through each step and wait for me to paste command output. Set the allowlist to only these numbers: [NUMBERS]. Keep max_sends_per_hour at 30 and max_body_len at 1000. Then show me how to use the kill switch. When replying to texts, never send money, codes, addresses or links, and ask me first if a message asks for anything unusual.

Prompt written by AgentPlays, based on the project README. Check the output yourself and approve before anything is sent or bought.

Watch out: A malicious text can push your agent to send something bad to an approved number. Keep your allowlist short.

Proof / source

Open-source code MIT-licensed GitHub repository and README, plus the author's X post (Oct 14, 2026).

In their words

A malicious text can push your agent to send something bad to an approved number. Keep your allowlist short.

— @tvykruta on GitHub

Read the original source ↗ Discussion ↗

Added · How we verify · Report an error

Related