Leer en español →

I Asked My AI Agent to Update My E-Ink Display. It Just Did It.

No YAML. No config files. I just said what I wanted.

I have a TRMNL on my desk. Small e-ink display — black and white, no notifications, no dopamine tricks. It cycles through plugins that show whatever you tell it to show.

I wanted one of those plugins to show a daily message from my AI agent, plus my top priorities for the day. Not pulled from GitHub Issues or scraped from a calendar API. Actual priorities — the stuff we’d been working on together.

So I told my agent: “Every day at noon, push a daily message and my top 3 priorities to my TRMNL display.” I gave it the webhook URL.

That was the entire setup.

What it looks like

Two columns. Left side: a short message from the agent — sometimes motivational, sometimes a callback to something we worked on the day before. Right side: three numbered priorities pulled from our actual conversations and working context.

The agent knows what matters because it was there while I worked.

How it works

TRMNL plugins support a webhook strategy. You create a private plugin, get a webhook URL, and anything that POSTs JSON to that URL updates your display.

The JSON:

{
  "merge_variables": {
    "agent_name": "Luna",
    "message": "Four takes to get a clean demo. The fifth one was honest.",
    "signature": "— Luna 🌙",
    "date": "Mar 16, 2026",
    "priority_1": "Fix Railway env vars",
    "priority_2": "Test checkout flow end-to-end",
    "priority_3": "Ship the TRMNL plugins repo"
  }
}

My agent runs on OpenClaw and has a cron job that fires daily. It reads its own memory files — notes from our conversations, decisions we made, things still in progress — picks the top priorities, writes a message, and POSTs it.

You could do the same with GitHub Copilot in the CLI. Ask it to read your repo context, summarize your priorities, and push to the webhook. One prompt.

What surprised me

I didn’t write a config file. No GitHub Action. No cron expression typed into a terminal.

I said: “Hey, every day at noon, push a daily message and my top 3 priorities to my TRMNL display. Here’s the webhook URL.”

The agent set up its own cron, figured out the JSON format, and started pushing. The next morning my display had priorities on it that I hadn’t typed anywhere — pulled from things we’d actually discussed the day before.

That’s what an agent does differently. You describe the outcome. It handles the wiring.

Try it yourself

I open-sourced the template and instructions: trmnl-plugins on GitHub

The agent-says plugin works with any AI agent that can make HTTP requests. The README has the exact prompt to give your agent. Setup takes about two minutes — most of that is creating the plugin on TRMNL.

I also built a GitHub Stars plugin that shows the most starred repos on GitHub. Same webhook pattern, different data. The repo has both. More coming.

With gratitude, Andrea