Thread Monitor

Activity monitor, but for your programming threads

A TUI for ML researchers and developers. Watch every training run, scraper, eval, worker, and long-lived process on your machine from one terminal pane. Built for the way real research work actually runs: long, parallel, and mostly happening over SSH.

Request early access Read the blog

thread-monitor

  PID     NAME                    STATE    CPU    MEM     RUNTIME    OWNER
  ─────   ─────────────────────   ──────   ────   ─────   ────────   ──────────
  48211   train_resnet50.py       RUN      312%   18.4G   04:12:37   torch
  48390   scrape_listings.py      RUN       42%    1.2G   22:18:04   playwright
  48441   eval_agent_suite.py     RUN      118%    6.8G   00:48:12   langgraph
  48502   ingest_pubsub_sink      IDLE       2%    0.3G   72:44:10   gcp
  48610   dev_server.next         RUN       18%    0.6G   01:22:09   node
  48712   fine_tune_qwen_2.5b     RUN      405%   31.2G   09:04:55   torch

  [tab] switch pane    [enter] inspect    [k] kill    [/] filter    [q] quit

Preview. Real screenshot landing soon.

What it does

Built for work that runs for hours.

Most developer tools assume you run a process, wait for it, then move on. ML work is the opposite. Thread Monitor is built for the other way around.

Every thread, one pane

See every training run, background worker, data job, and long-lived process your machine is running. No tab switching, no hunting for a PID, no guessing which terminal owns what.

Built for long-running work

ML work lives on the clock. Training runs for hours, scrapers for days, evals overnight. Thread Monitor is built for processes you started yesterday and forgot the name of.

Lives where you work

It's a TUI. It opens in the same terminal you ssh into, the same tmux session you left running on the GPU box, the same place you already live. No browser, no electron, no context switch.

Fast and quiet

Low overhead, no dashboards to babysit. It sits in a pane, updates on its own, and stays out of your way until you need it.

Who it's for

ML researchers and developers.

If you've ever lost a training run because you forgot which tmux pane it was in, this is for you.

→ML researchers juggling multiple training runs across local and remote machines
→AI engineers running evals, fine-tunes, and agent experiments in parallel
→Data engineers with long-lived scrapers, ingestion jobs, and pipelines
→Developers who live in the terminal and want activity monitor ergonomics without leaving it

FAQ

Questions we get.

What is Thread Monitor?

Thread Monitor is a TUI that shows every programming thread and long-running process on your machine in one place. Think activity monitor, but built for the way researchers and developers actually work: in the terminal, across long-running jobs, without the browser or electron overhead.

Who is it for?

ML researchers, AI engineers, and developers who run multiple long-lived processes in parallel. If you have three training runs, a scraper, a dev server, and an eval queue going at once, Thread Monitor is built for you.

Why a TUI and not a web dashboard?

Because the work already happens in the terminal. A TUI opens where you ssh into, fits into tmux, survives disconnects, and doesn't ask you to switch context into a browser tab. The fewer tools between you and the process, the better.

Is it open source? How do I get it?

We're preparing a public release. If you're an ML researcher or developer in Australia and want early access, get in touch and we'll add you to the first cohort.

Want early access?

Thread Monitor is in private beta with a small group of researchers and engineers. Get in touch if you want to be in the next cohort.

Request access