liamsysmind

liamsysmind / roost

Public

A self-hosted workspace where your AI agents come to roost.

14
1
100% credibility
Found May 19, 2026 at 14 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Go
AI Summary

Roost is a self-hosted browser workspace that lets you access a full terminal on your Linux development machine from any web browser on any device. You install one small program on your Linux box, connect through a secure SSH tunnel, and open a browser to get a real terminal, file browser, and AI activity panel all in one page. Your work survives disconnects, network changes, and even computer restarts because it runs on tmux underneath. There's no cloud subscription, no account needed—just one binary, one password, and your browser.

How It Works

1
💡 You discover roost

You hear about a tool that lets you access your Linux development machine from any browser on any device.

2
🖥️ You install it on your Linux machine

You download one small program and install it on your development computer in just a few minutes.

3
🔐 You set a password

The program asks you to create a password to keep your workspace private and secure.

4
🔌 You connect through SSH

You run one simple command to create a private tunnel from your laptop to your development machine.

5
🌐 You open any browser

You type a local web address, enter your password, and suddenly your terminal appears in the browser.

6
You explore your workspace
⌨️
Using the terminal

Type commands, run builds, and watch your AI coding assistant work in real-time.

🤖
Watching AI activity

See every prompt you sent to your AI assistant, click any one to jump back to that moment.

Everything just works

Your terminal survives network hiccups, you can open it on your phone or tablet, and your work is always there when you return.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 14 to 14 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is roost?

Roost is a self-hosted browser-based workspace for terminal-first developers who work with AI coding agents. You install a single Go binary on a Linux server, SSH-tunnel into it, and access everything through a web browser. The interface gives you a persistent tmux-backed terminal that survives network disconnects, a file browser with inline previews for code and media, and an AI activity panel that reads your Claude Code or Codex session logs to show prompts and context tokens. Each browser tab becomes its own named shell session with a unique URL, so closing your laptop and reconnecting later replays your full scrollback. Configuration is stored in TOML, authentication is password-based, and the whole thing stays on your own machine with no cloud dependency.

Why is it gaining traction?

The core appeal is session persistence without the cloud. Your terminal survives laptop sleep, network changes, and server restarts because tmux holds the shell state. When you reconnect, your scrollback is there. The AI activity panel is the differentiator for Claude Code and Codex users: it surfaces prompts and token counts from the agent's own session logs, letting you jump back to specific AI interactions in the terminal scrollback. File operations work without leaving the browser—drag-drop uploads stream directly to disk with no size caps, and previews render markdown, code, images, and PDFs inline. The SSH-tunnel deployment model means it works on any OS with zero install on your laptop.

Who should use this?

Backend and DevOps developers who run AI coding agents (Claude Code, Codex) on a remote Linux box and want a unified browser workspace without cloud subscriptions. It's particularly useful if you frequently switch between machines or networks and need your terminal sessions to persist. If you want VS Code in a browser, look elsewhere—but if you want a resilient, self-hosted terminal that knows about your AI agent's conversation history, this fills a specific gap. Multi-user scenarios require separate instances per person by design.

Verdict

The concept is solid and the implementation is thoughtful, but the credibility score of 1.0% and 14 stars reflect a very early-stage project with minimal community validation. The documentation is thorough and the feature set is coherent, but without a 1.0 release or broader adoption, production use carries unquantified risk. Try it on a personal dev box if the use case fits—you'll know within an hour whether it works for your workflow.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.