xbot

xbot is a Go framework for building AI agents. It provides a message bus + plugin architecture where an Agent (LLM + tools + memory) receives messages from any Channel (CLI, Feishu, QQ, Web) through a Bus, processes them in a multi-turn loop with tool calling, and sends replies back. Designed for self-hosted deployments, it supports OpenAI and Anthropic as native LLM providers, plus any OpenAI-compatible API (DeepSeek, Qwen, Ollama, etc.) via the openai provider with a custom base_url.

┌──────────┐     ┌──────────────┐     ┌────────┐     ┌──────────┐
│  Feishu  │────▶│  Dispatcher  │────▶│ Agent  │────▶│   LLM    │
│  QQ      │◀────│  (channel/)  │◀────│ (agent/)│◀────│ (llm/)   │
│  NapCat  │     └──────────────┘     │        │     └──────────┘
│  Web     │                          │        │
│  CLI     │                          │        │────▶ Tools
└──────────┘                          │        │      (tools/)
                                      │        │
                                      │        │────▶ Memory
                                      │        │      (memory/)
                                      └────────┘

# Default: installs xbot-cli to /usr/local/bin
curl -fsSL https://raw.githubusercontent.com/CjiW/xbot/master/scripts/install.sh | bash

# Specific version
VERSION=v0.0.7 curl -fsSL https://raw.githubusercontent.com/CjiW/xbot/master/scripts/install.sh | bash

# Custom install path
INSTALL_PATH=~/.local/bin curl -fsSL https://raw.githubusercontent.com/CjiW/xbot/master/scripts/install.sh | bash

git clone https://github.com/CjiW/xbot.git && cd xbot
make build          # Builds xbot (server + runner)
make run            # Build and run server

To build xbot-cli only:

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

• Multi-channel — Pluggable channel adapters: CLI (TUI), Feishu (Lark), QQ, NapCat (OneBot 11), Web
• Tools — 50+ built-in tools: Shell, File I/O, Web fetch/search, Context editing, SubAgent, Cron scheduling, Download, Feishu MCP, and more
• Memory — Pluggable providers: Flat (in-memory blocks + grep archival) and Letta/MemGPT (SQLite core + vector search + FTS5)
• Skills & Agents — Markdown-defined skill packages; role-based SubAgents with custom roles, max nesting depth 6
• MCP Protocol — Global and session-scoped MCP servers, stdio and HTTP transports, lazy cleanup
• Permission Control — OS user-based permission control with approval workflows for privileged operations
• Multi-tenant — Channel + chatID isolation
• OAuth 2.0 — Built-in OAuth server for web channel authentication
• Hot-reload prompts — Go templates with channel-specific overrides
• KV-Cache optimized — Context ordering maximizes LLM cache hits

Each channel is a pluggable adapter on the message bus. See the Channels page for setup guides and configuration details.