I made ATM, short for Agent Task Manager. It is a small CLI/TUI that shows which local coding agents are running on your machine, which repo they are in, and whether they look active, idle, stale, or just present as a process.
It currently recognizes common agent CLI processes like Codex, Claude, Gemini, Aider, OpenCode, Goose, Amp, and Cursor Agent. Codex has the deepest support right now: ATM can read local session metadata to show the project, session path, last activity, and a compact "what just happened" summary.
I built it because I kept ending up with several agents running across different repos and wanted something friendlier than ps for answering: "what is still alive, where is it working, and has it gone quiet?"
I made ATM, short for Agent Task Manager. It is a small CLI/TUI that shows which local coding agents are running on your machine, which repo they are in, and whether they look active, idle, stale, or just present as a process.
It currently recognizes common agent CLI processes like Codex, Claude, Gemini, Aider, OpenCode, Goose, Amp, and Cursor Agent. Codex has the deepest support right now: ATM can read local session metadata to show the project, session path, last activity, and a compact "what just happened" summary.
I built it because I kept ending up with several agents running across different repos and wanted something friendlier than ps for answering: "what is still alive, where is it working, and has it gone quiet?"
Install:
curl -fsSL https://github.com/artpar/atm/releases/latest/download/insta... | sh
Repo: https://github.com/artpar/atm
It is early, tiny, and CLI-first. I’d love feedback from people running multiple local agents at once because right now its a very basic.