avibe-os 3.0.0

Local-first Agent OS for Web and IM agent workflows

Avibe

Your AI colleagues, present in Slack, Discord, Telegram, WeChat & Lark.

No laptop. No IDE. Just vibes.

Docs | English | 中文

Supported Platforms

Supported Agents

---

The Pitch

You're at the beach. Phone buzzes — production's on fire.

Old you: Panic. Find WiFi. Open laptop. Wait for IDE. Lose your tan.

With Avibe: Open Slack, Discord, Telegram, or WeChat. Type "Fix the auth bug in login.py". Watch Claude Code fix it in real-time. Approve. Sip margarita.

AI works. You live.

---

Install in 10 Seconds

curl -fsSL https://avibe.bot/install.sh | bash && vibe

Open source — view the script on GitHub. The short URL is a 307 redirect to that file.

That's it. Browser opens -> Follow the wizard -> Done.

Need to open the Web UI from another device or a remote server? Run:

vibe remote

Windows?

We recommend the WSL setup on Windows because it has the best compatibility.

• Recommended: Run Avibe with WSL from Scratch

If you are new to WSL, this guide explains:

• where to install WSL
• which terminal window to use
• where to run the Avibe install command
• how to launch Ubuntu and open the Web UI

---

What Ships Today

Not an agent framework. Not a hosted coding VM. Avibe is the local-first Agent OS that brings the AI agents already running on your machine into the Web and chat surfaces where you already collaborate.

• Chat apps become collaboration spaces: Slack, Discord, Telegram, WeChat, and Lark / Feishu.
• Real agents, no middleman: Today's first-class backends are Claude Code, OpenCode, and Codex.
• Thread = session: Start five threads, run five isolated agent jobs, resume later.
• Web setup, not token archaeology: Local wizard, dashboard, routing, and health checks.
• Remote UI when you need it: vibe remote opens your local Web UI through a secure avibe.bot tunnel.
• Walk away without losing the loop: Completion notifications and the Agent Harness keep work moving while you live your life.

---

Why This Exists

Problem	Solution
Claude Code is amazing but lives in a terminal	Slack/Discord/Telegram/WeChat/Lark can become the collaboration surface
Context-switching kills flow	Stay in one app
Start on desktop, continue on phone	Resume the exact agent session from the current project in seconds
Can't code from phone	Yes you can
Multiple agents, multiple setups	One chat app, any agent

Supported Agents:

• Claude Code — Deep reasoning, complex refactors
• OpenCode — Fast, extensible, community favorite
• Codex — OpenAI's coding model

---

Meet Vibey

Lives in your Slack, Discord, Telegram, Lark, or WeChat. Reads the room. Picks up where you left off. Asks the right question when it's not sure. Goes quiet when you're heads-down. Ships at 2am because that's when the vibe hits — then leaves a note about what it touched.

Avibe is the wire. Vibey is the colleague on the other end.

Forgets nothing. Holds opinions. Says thanks when you fix its bugs.

---

Why Avibe over OpenClaw?

	Avibe	OpenClaw
Setup	One command + web wizard. Done in 2 minutes.	Gateway + channels + JSON config. Expect an afternoon.
Security	Local-first. Socket Mode / WebSocket only. No public endpoints, no inbound ports, minimal attack surface.	Gateway exposes ports. More moving parts, more attack surface.
Token cost	No extra reasoning loop in the middleware. Tokens go to your chosen agent, not to a second assistant layer.	Every message carries a long system context for maintaining agent persona, IM tooling, and orchestration plumbing. Tokens burn on overhead before your actual task even starts.

OpenClaw is a personal AI assistant — great for casual chat, but its always-on agent loop makes it expensive for real productivity workloads. Avibe is not an agent framework. It is a local-first collaboration runtime for the AI agents you already trust: the agent stays itself, your data stays local, and the colleague experience comes from putting the agent into the same communication flow where human work already happens. Coding is the first strong workload, not the product boundary. Every token goes straight to your task.

---

Highlights

Setup Wizard One-command install, guided configuration. No manual token juggling.

Dashboard Real-time status, health monitoring, and quick controls.

Channel Routing Per-channel agent configuration. Different projects, different agents.

Instant Notifications

Get notified the moment your AI finishes. Like assigning tasks to employees — delegate, go do something else, and come back when the work is done. No need to babysit.

Thread = Session

Each Slack/Discord/Telegram/WeChat/Lark chat scope is an isolated workspace. Open 5 chats, run 5 parallel tasks. Context stays separate.

Resume Anywhere

Laptop closed. Commute started. Production still needs you.

Avibe lets you reopen real agent sessions from your current working directory across Claude Code, OpenCode, and Codex. Pick the latest session, jump from desktop to mobile, and keep going without re-explaining the task, hunting for an old thread, or losing the thread of thought.

Interactive Prompts

When your agent needs input — file selection, confirmation, options — your chat app pops up buttons or a modal. Full CLI interactivity, zero terminal required.

---

Harness

Avibe gives agents a real harness: a small set of durable tools they can combine to work on their own timeline, more like a colleague than a command you have to babysit.

Think of it like giving your AI colleague a task board, a calendar, and a notification center:

• vibe agent run starts one concrete Agent Run. It can be synchronous when you want an answer now, or async when the agent should keep working in the background.
• vibe task saves time-based work: "run this every morning", "remind me tonight", "check this once after lunch".
• vibe watch waits for the world to change: a PR review, CI result, deployment, file, log line, or long-running process.
• vibe runs lets the agent inspect what happened, continue from a run, or cancel work when the situation changes.

The important part is that you do not have to memorize all of that. Agents can learn these primitives and use them as building blocks.

Ask in normal language:

• "Watch this PR and come back when there is actionable review feedback."
• "Run this deployment check every weekday morning and post the summary here."
• "Start a separate investigation session for this incident, but report the conclusion to this channel."
• "Kick off three background checks and keep track of their run IDs."
• "If CI fails, summarize the logs; if it passes, tell me whether the PR is mergeable."

That is the Harness: the agent can leave the current chat turn, wait, return, branch, retry, and keep records without inventing fragile shell scripts or making you poll manually.

---

📡 Your Machine, In Your Pocket — Powered by avibe.bot

Your Avibe runs on your machine. You don't sit in front of it 24/7.

You're on a plane. At a café. Holding someone else's borrowed laptop. The bot just pinged you that a job needs attention — and the control panel is back home.

vibe remote

One command. Your local Web UI becomes reachable from any browser on Earth — through a secure avibe.bot tunnel.

• 🌍 Your own you-app.avibe.bot — 30-second sign-in, your slug for life
• 🔒 Fail-closed at every join — Auth, routing, host checks default to "deny"
• 📱 Mobile-aware UI — Thumb-friendly layout, designed for borrowed screens
• ⏱ 24-hour sessions — Cookie auto-renews mid-session, no sudden logouts

No VPN. No port forwarding. No "wait what's my IP again." No public webhooks pointed at your laptop. Your data plane stays on your machine; avibe.bot is just the control-plane handshake.

Your agents work on your machine. You command them from your pocket.

→ Set it up in 60 seconds

---

How It Works

┌──────────────┐             ┌──────────────┐             ┌──────────────┐
│     You      │   Slack     │              │   stdio     │  Claude Code │
│  (anywhere)  │   Discord   │    Avibe     │ ──────────▶ │  OpenCode    │
│              │   Telegram  │ (your host)  │ ◀────────── │  Codex       │
│              │   WeChat    │              │             │              │
│              │   Lark      │              │             │              │
└──────────────┘             └──────────────┘             └──────────────┘

1. You type in Slack/Discord/Telegram/WeChat/Lark: "Add dark mode to the settings page"
2. Avibe routes to your configured agent
3. Agent reads your codebase, writes code, streams back
4. You review in your chat app, iterate in thread

Your data never leaves your machine. Avibe runs locally and connects via Slack Socket Mode, Discord Gateway, Telegram Bot API long polling, WeChat polling, or Lark WebSocket.

---

Commands

In chat	What it does
@avibe /start	Open control panel
/stop	Kill current session
Just type	Talk to your agent
Reply in thread	Continue conversation

Pro tip: Each thread = isolated session. Start multiple threads for parallel tasks.

Detailed references:

• Full Command Reference
• CLI Reference

---

Instant Agent Switching

Need a different agent mid-conversation? Just prefix your message:

Plan: Design a new caching layer for the API

That's it. No menus, no commands. Type AgentName: and your message routes to that agent instantly.

---

Per-Channel Routing

Different projects, different agents:

#frontend    → OpenCode (fast iteration)
#backend     → Claude Code (complex logic)  
#prototypes  → Codex (quick experiments)

Configure in web UI → Channels.

---

CLI

vibe          # Start everything
vibe status   # Check if running
vibe stop     # Stop everything
vibe doctor   # Diagnose issues
vibe task     # Create and manage scheduled tasks
vibe watch    # Manage background waiters
vibe agent    # Manage and run Avibe Agents
vibe runs     # Inspect Agent Run records

Detailed references:

• Full Command Reference
• CLI Reference

---

Prerequisites

You need at least one coding agent installed:

OpenCode (Recommended)

curl -fsSL https://opencode.ai/install | bash

Required: Add to ~/.config/opencode/opencode.json to skip permission prompts:

{
  "permission": "allow"
}

Claude Code

npm install -g @anthropic-ai/claude-code

Codex

npm install -g @openai/codex

---

Security

• Local-first — Avibe runs on your machine
• Socket Mode / WebSocket — No public URLs, no webhooks
• Your tokens — Stored in ~/.avibe/, never uploaded. Existing installs keep ~/.vibe_remote/ as a compatibility path.
• Your data — Stays on your disk, sent only to your chosen AI provider

---

Uninstall

vibe stop
uv tool uninstall avibe-os
uv tool uninstall vibe-remote
rm -rf ~/.avibe ~/.vibe_remote

---

Roadmap

• Slack support
• Discord support
• Telegram support
• WeChat support
• Lark (Feishu) support
• Web UI setup wizard & dashboard
• Per-channel agent routing
• Interactive prompts (buttons, modals)
• File attachments
• SaaS Mode
• Avibe Coding Agent (one agent to rule them all)
• Skills Manager
• Best practices & multi-workspace guide

---

Docs

• Official Docs — Quickstart, platform guides, agent setup, troubleshooting, and AI-readable install instructions
• CLI Reference — Command-line usage and service lifecycle
• Install via AI Agent — Give this to Claude Code, Codex, or OpenCode for guided setup
• Incus Tenant Scaffold — Create isolated Avibe tenants on one Incus host
• Slack Setup Guide — Detailed setup with screenshots
• Discord Setup Guide — Detailed setup with screenshots
• Telegram Setup Guide — BotFather setup, token validation, and chat discovery
• WeChat Setup Guide — Follow the in-app wizard (vibe → choose WeChat)
• Lark Setup Guide — Follow the in-app wizard (vibe → choose Lark)

---

Stop context-switching. Start vibe coding.

Install Now · Setup Guide · Report Bug · Follow @alex_metacraft

---

Built for developers who code from anywhere.