Product Thumbnail

nlsh

talk to your terminal in natural language

Open Source
Languages
Developer Tools
GitHub

A terminal interface that translates plain English into shell commands. Stop memorizing flags. Just type what you want.

Top comment

Good idea, but once installed you can't use the terminal across the entire system anymore and you also can't stop the app. That's kind of weird.

Comment highlights

Congrats! Awesome product!
I'm curious how do you differentiate from warp?

As someone who spends most of my day in Nushell and Kitty, I’m curious—how does nlsh handle complex piped commands? Does it 'suggest' the syntax so I can learn the raw command, or does it just execute the abstraction? Very cool for speeding up DevOps tasks.
This is awesome! The main reason I use Warp is for the natural language AI conversions. Definitely going to check this out.

It is Gemini-based. But if I have Gemini, I can use Gemini CLI. What feature do you have, and how is it better than Gemini CLI?

This looks cool! I’m building a similar system for myself that monitors things like dmesg/journalCTL logs passively and scans installed packages for man files that it uses as sources of truth for writing commands. Really helps me utilize tools like FFMPEG and imagemagick to their full capacity. Worth looking into if you want to build this out more

@jun3id That's a cool idea! It makes sense to use AI for terminal commands since so many of them are very hard to memorise. A bit scary though for things like killing processes, moving/removing files and mostly git commands 🤣 Are there any guardrails?

How have you designed the review/confirmation flow so it actually reduces cognitive load instead of forcing users to become “auditors” of every command—especially for destructive or multi-step operations?

Oh wow. So if I install a tool and ask this app on my terminal to run a command , will it get me the working command ?

This looks really cool. Would be awesome to see if you could run it off a local model too. Perhaps a small one only suited for this task.

I don't know how many turns it keeps in context window, I'm curious.