Installation

# Linux / macOS (amd64, arm64)
curl -fsSL https://raw.githubusercontent.com/ai-pivot/xbot/master/scripts/install.sh | bash

# Windows (PowerShell)
irm https://raw.githubusercontent.com/ai-pivot/xbot/master/scripts/install.ps1 | iex

Pin a version or choose an install path:

VERSION=v0.0.48 curl -fsSL ... | bash            # specific version
INSTALL_PATH=~/.local/bin curl -fsSL ... | bash  # custom path

Note

Behind a firewall (China)? Use the mirror-accelerated installer — it auto-detects a working CDN mirror and proxies all GitHub downloads:

curl -fsSL https://ghfast.top/https://raw.githubusercontent.com/ai-pivot/xbot/master/scripts/install-cn.sh | bash

git clone https://github.com/ai-pivot/xbot.git && cd xbot
make build          # build xbot (server + runner)
make run            # build and run the server

Build only the CLI:

go build -o xbot-cli ./cmd/xbot-cli

Requirements: Go 1.26+. The Web UI bundles are committed, so no Node.js is needed to build the Go binaries.

The installer lets you choose Standalone or Server.

The CLI runs the agent locally with no background service.

• ✅ Simple, install-and-go
• ✅ No background process
• ❌ Stops when you close the terminal
• ❌ CLI channel only — no Feishu / QQ / Web
• ❌ No team-shared LLM

Best for: solo developers who want a quick test drive.

A background server process runs continuously. CLIs connect over WebSocket, and Feishu / QQ / Web channels are enabled simultaneously.

• ✅ Agent runs 24/7, auto-starts on boot
• ✅ Feishu / QQ / Web channels all active
• ✅ Web browser chat UI
• ✅ Admin configures the LLM key once — the whole team uses it
• ✅ Multiple CLI clients connect at once

Best for: teams, anyone who needs Feishu / QQ / Web, or wants a Web UI.

Important

Most teams should choose Server mode.

The installer configures a user-level system service (no sudo needed):

Start the server with: xbot-cli serve

1. Downloads xbot-cli to ~/.local/bin/ (or your custom path)
2. Generates a random admin token
3. Writes / updates ~/.xbot/config.json
4. Server mode: installs a system service + downloads the Web UI

After installing, run:

xbot-cli

The first run auto-launches the Setup wizard, which guides you through:

LLM subscription

1. Choose a provider (OpenAI / Anthropic / OpenAI-compatible API)
2. Enter your API key (required)
3. Set the API base URL (default https://api.openai.com/v1; change for compatible services)
4. Choose a model
5. Configure model tiers (Vanguard / Balance / Swift)
6. Tavily search key (optional — enables web search)

Environment

• Sandbox mode (default none; Docker users choose docker)
• Memory provider (default flat)

Appearance

• Color scheme (9 built-in themes)

Re-run anytime via /setup or Ctrl+K → Setup.

You can also edit ~/.xbot/config.json directly. See Configuration reference for all fields.

{
  "subscriptions": [
    {
      "name": "default",
      "provider": "openai",
      "api_key": "sk-xxx",
      "model": "gpt-4o"
    }
  ]
}

{
  "subscriptions": [
    {
      "name": "DeepSeek",
      "provider": "openai",
      "api_key": "your-key",
      "base_url": "https://api.deepseek.com/v1",
      "model": "deepseek-chat"
    }
  ]
}

xbot-cli --version

# Server mode — check service status
# Linux:
systemctl --user status xbot-server
# macOS:
launchctl list | grep xbot

Tip

Quick health check: Run xbot-cli and type “hello”. If the agent responds, everything is working correctly.

• Getting Started — 5-minute quick start
• Configuration — all config.json fields
• Channels — Feishu, QQ, Web, CLI setup