TaskYouOS

Your own AI agent team, running on a remote server and managed through Claude Code.

You describe what you need in plain English. A "General Manager" running in Claude Code delegates work to background agents on a remote server. They research, analyze, write, and build — even while your laptop is closed.

Get Started

You need Claude Code. Open it and run:

/plugin marketplace add taskyou/taskyou-os
/plugin install taskyou-os
/taskyou-os:launch

That's it. The setup walks you through everything:

1. What's your project? — Tell it what you're working on and it designs your agent workspaces
2. Server — Spins up a cloud server on exe.dev (or uses your own). Installs everything automatically
3. Config — Writes your configuration file based on your answers
4. Deploy — Sets up the server, authorizes your agents, starts the daemon
5. Done — Gives you a launch command and shows you how to use your new GM

No git cloning, no manual config editing, no SSH knowledge required. The setup handles SSH keys, exe.dev signup, software installation, and authentication transfer for you.

You can re-run /taskyou-os:launch anytime to resume an interrupted setup or fix issues.

How It Works

Your Machine                      Server
┌──────────────┐                  ┌──────────────────┐
│ Claude Code  │ ──── SSH ─────→  │ TaskYou daemon   │
│ (GM session) │                  │ ┌──────────────┐ │
│              │                  │ │ Agent 1      │ │
│              │                  │ │ Agent 2      │ │
│              │                  │ │ ...          │ │
│              │                  │ └──────────────┘ │
│              │                  │                  │
│              │                  │ notifications    │
└──────────────┘                  │ .jsonl           │
                                  └──────────────────┘

• You talk to the GM (a Claude Code session on your machine) in plain English
• The GM creates tasks and assigns them to agents on the server
• Agents work in the background — close your laptop, they keep going
• The GM automatically tracks running tasks and notifies you when they complete or get blocked — no need to remember to check
• Use /gm-babysit for an immediate status check on all tracked tasks

exe.dev Deployment (No Local Machine Required)

Instead of running the GM locally over SSH, you can deploy it entirely onto an exe.dev VM. This gives you a web-accessible GM with:

• A landing page with a live kanban board
• A one-click terminal button that opens the GM in your browser
• No local machine required — access from anywhere

Browser                           exe.dev VM
┌──────────────┐                  ┌──────────────────┐
│ Landing page │ ──── HTTPS ───→  │ nginx (port 8000)│
│ (kanban)     │ ← board.json ──  │                  │
│              │                  │ TaskYou daemon   │
│ GM Terminal  │ ──── xterm ───→  │ ┌──────────────┐ │
│ (browser)    │                  │ │ Agent 1      │ │
│              │                  │ │ Agent 2      │ │
│              │                  │ │ ...          │ │
│              │                  │ └──────────────┘ │
└──────────────┘                  └──────────────────┘

To deploy:

# Set EXE_DEV_VM_NAME in your config.env, then:
./setup.sh exe ~/Projects/gms/myproject

This creates the VM, uploads everything, and prints the URL. Share access with teammates via ssh exe.dev share add <vm> user@example.com.

GM Commands

Once your GM is running, these commands are available:

Command	What it does
/gm-babysit	Immediate status check on all tracked tasks
/gm-status	Daemon, tasks, and agents overview
/gm-fix	Diagnose and fix agent system problems
/gm-start	Getting started guide with project-specific examples
/gm-help	Quick reference for all commands and TaskYou CLI

Plugin Commands

These run from the taskyou-os plugin context (not inside a GM):

Command	What it does
/taskyou-os:launch	Interactive wizard to create a new GM
/taskyou-os:doctor	Health check — updates plugin, checks versions, daemon status, executor health, template drift, security audit

Run /doctor periodically to keep things healthy. It also detects new commands available from the plugin and offers to add them to existing GMs.

Updating

Update the Plugin

The taskyou-os plugin updates through the Claude Code marketplace. Run:

/plugin marketplace add taskyou/taskyou-os

This pulls the latest version. Restart Claude Code afterward for changes to take effect.

Update a GM Installation

Run /taskyou-os:doctor from any directory. It checks and updates everything automatically:

1. Plugin version — pulls the latest from the marketplace and updates the local cache
2. TaskYou binary — upgrades ty locally and on your server via ty upgrade
3. Daemon health — verifies the daemon is running, restarts it if needed
4. Executor health — confirms every active task has a running executor pane, recovers orphaned tasks
5. Command migration — removes old local command files that shadow newer plugin-delivered versions
6. CLAUDE.md drift — detects new sections in the plugin template and offers to add them to your GM
7. Security audit — runs the server-side credential and permissions check
8. Credential isolation — verifies nono sandbox setup and detects template drift

If /doctor finds issues, it fixes what it can and tells you what to do for the rest.

Update the Server

If you need to re-deploy server-side files (hooks, scripts, nono config) after a plugin update:

./setup.sh server ~/Projects/gms/myproject

This re-renders templates from your config.env and uploads them. It won't touch your data or running tasks.

Built-in Modules

These are configured via flags in config.env during setup. They're part of the repo, not separate installs.

• Linear (LINEAR_ENABLED=true) — Agent-to-human handoff via Linear issues, plus @agent comments for revisions
• Cloudflare R2 (R2_ENABLED=true) — Public URLs for files and assets agents generate
• GitHub (GITHUB_REPOS=workspace:org/repo) — Push agent work to your repositories
• nono (NONO_ENABLED=true) — Credential isolation for agents via sandboxed executor wrappers

Manual Setup

If you prefer to skip the interactive /launch wizard:

mkdir -p ~/Projects/gms/myproject
cp config.example.env ~/Projects/gms/myproject/config.env
# Edit config.env with your values
./setup.sh all ~/Projects/gms/myproject

Modes: local (local machine only), server (remote server only), exe (exe.dev VM), all (local + server).

See config.example.env for all available configuration options.