Post History

Current version by Nick Antonaccio

Current VersionSep 28, 2026 at 18:23

I described pi-web to a colleague like this:

This is getting to be my favorite interface for all my LLM work:

https://github.com/agegr/pi-web

It's such a simple tool, but is quickly earning a position as the central hub of every bit of work I do. Pi-web connects to Pi wherever you choose to install it, and provides a web interface to all of Pi's features. At first it looks like another locally hosted ChatGPT/Claude chat clone, but it enables the full agentic power of Pi, to build and deploy software directly on your server, control your hardware/OS/files/software/etc., in any way you need, using any LLMs you choose (locally hosted or from any API provider - anything you can connect Pi to, all hot swappable), directly from any browser, on any client device, from any location.

Anything you do in the pi-web browser app is shared 1-1 with anything you do directly in Pi at the command line of your server computer (you control all your saved sessions and every other piece of your own installed Pi configuration, with all the same detail available in a console). The UI is laid out nicely, anyone can use it, and it takes just a few minutes to install (5 pasted lines at an SSH console, plus a Caddy entry).

Set up Pi and pi-web on a VPS, attach Cline-Pass, OpenRouter, or any other LLM API, and/or any self-hosted LLMs, and you can orchestrate every bit of work you do, from a phone or any other machine that has even a very old browser. I have it running perfectly smoothly on 2 ancient Amazon Fire devices I bought at a garage sale for $5 :) I can control my entire kingdom with it, and non-technical clients can use it immediately to accomplish just about anything... It feels just like using ChatGPT, no training required, to integrate LLM usage on just about any mix of of local computer hardware, OSs, remote server setups, and client devices which anyone might choose to use. Damn cool.

Previous Versions
Version 1Sep 28, 2026 at 18:23

I described pi-web to a client like this:

This is getting to be my favorite interface for all my LLM work: https://github.com/agegr/pi-web

It's such a simple tool, but is quickly earning a position as the central hub of every bit of work I do. Pi-web connects to Pi wherever you choose to install it, and provides a web interface to all of Pi's features. At first it looks like another locally hosted ChatGPT/Claude chat clone, BUT it enables the full agentic power of Pi, to build and deploy software directly on your server, control your hardware/OS/files/software/etc., in any way you need, using any LLMs you choose (locally hosted or from any API provider - anything you can connect Pi to, all hot swappable), directly from any browser, on any client device, from any location.

Anything you do in the pi-web browser app is shared 1-1 with anything you do directly in Pi at the command line of your server computer (you control all your saved sessions and every other piece of your own installed Pi configuration, with all the same detail available in a console). The UI is laid out nicely, anyone can use it, and it takes just a few minutes to install (5 pasted lines at an SSH console, plus a Caddy entry).

Set up Pi and pi-web on a VPS, attach Cline-Pass API or any self-hosted LLMs, and you can orchestrate every bit of work you do, from a phone or any other machine that has even a very old browser. I have it running perfectly smoothly on 2 ancient Amazon Fire devices I bought at a garage sales for $5 :) I can control my entire kingdom with it, and non-technical clients can use it immediately to accomplish just about anything... It feels just like using ChatGPT, no training required, to integrate LLM usage on just about any mix of of local computer hardware, OSs, remote server setups, and client devices which anyone might choose to use. Damn cool.

Back to Thread