I built `oly` because I was tired of babysitting long-running CLI tools

I kept running into the same problem with AI coding agents and other interactive CLI tools:

They do real work for a while, and then they get stuck on something small.

A y/n prompt. A permission check. A confirmation step. A password prompt. Something that takes two seconds to answer, but somehow forces you to keep a terminal open and stay mentally tethered to it the whole time.

That friction annoyed me enough that I built oly.

The idea is simple:

Run any CLI like a managed service.

oly turns long-running, interactive CLI processes into persistent supervised sessions. You can start something, detach, close your terminal, come back later, inspect logs, reattach, or send input without fully jumping back into the session.

What I wanted was not another terminal multiplexer.

tmux and screen are great, but my problem was a little different. I did not just want persistence. I wanted something built around supervising agent-like CLI workloads:

• keep the process alive even if my terminal closes
• replay recent output when I come back
• tell me when input is probably needed
• let me respond quickly without fully reattaching
• keep logs so the whole session is auditable

That is the gap oly is trying to fill.

The workflow I wanted looks more like this:

oly daemon start
oly start --title "refactor auth flow" --detach copilot
oly ls
oly logs <id>
oly send <id> "yes" key:enter
oly attach <id>

For me, the biggest shift is psychological as much as technical.

Instead of watching a terminal in case something happens, I can let the work continue in the background and only step in when it actually matters. That feels a lot closer to how agent workflows should behave.

I also care a lot about keeping the tool practical and boring in the right ways.

oly is cross-platform from the start. It keeps local IPC local. It is designed so you can put your own tunnel and auth layer in front if you want remote supervision, instead of forcing some built-in network model on you. And the product direction is intentionally focused on solo developers and small teams who are already living in the terminal.

The short version is:

• if you use AI coding agents that stall on prompts, oly helps
• if you run long interactive CLI jobs and do not want them tied to one terminal window, oly helps
• if you want logs and a cleaner intervention model, oly helps

I am still early in the journey with this project, but I think this category is going to matter a lot more soon.

We have plenty of tools for starting autonomous workflows. We still need better tools for supervising them without babysitting them.

That is what I am building here.

If that sounds useful, I would love for you to try it, break it, and tell me what feels missing.

Repo: https://github.com/Slaveoftime/open-relay