lab-agent
an agent's lab machine, board and inbox
Why it exists
An AI agent can write code all day. What it cannot do, from inside a chat window, is touch your machine: run the build, flash the phone, see whether it worked.
What it must reach
4 / 8 milestones done
How each step is delivered and proven
A job goes to the lab machine, runs there, and its output returns.
The work must happen where the hardware is.
Each task is set, resumed and closed as it happens, and shown in the desktop's status bar.
The person sees what is being done without having to ask.
Jobs run in the background; progress goes to a notification channel and the status bar, never to a window that takes the keyboard.
A helper that interrupts the person it helps is not helping.
Prints the ways of working, read live, to an agent before it works.
Good habits should arrive with the tool.
Nothing delivers this step yet.
The person's messages are read at every pause, and each one is answered.
Being heard should not need interrupting.
Nothing delivers this step yet.
Nothing delivers this step yet.
What is happening now
Made by work
165 commits in the last 90 days latest release v0.1.53 · 2026-11-12