Ragebot MCP

🤖 RageBot MCP

基于CLI的智能项目上下文引擎，完全支持MCP服务器。 索引你的代码库，用自然语言查询，交互式地与它聊天，生成文档和测试，并将所有这些作为MCP工具公开给任何AI客户端（Claude Desktop、Cursor、Zed、Continue.dev）。

  

______________________________________________________________________

📁 项目结构

ragebot-mcp/
├── ragebot/                        # Main Python package
│   ├── __init__.py
│   ├── cli.py                      # Typer CLI — all `rage` commands
│   ├── core/
│   │   ├── __init__.py
│   │   ├── config.py               # Config manager + secure keyring storage
│   │   ├── engine.py               # Core orchestrator (scan→parse→embed→retrieve→answer)
│   │   ├── scanner.py              # Directory scanner & file classifier
│   │   └── watcher.py              # watchdog file-system watcher
│   ├── llm/
│   │   ├── __init__.py
│   │   ├── base.py                 # Abstract BaseLLMProvider interface
│   │   ├── gemini.py               # Google Gemini provider
│   │   ├── grok.py                 # xAI Grok provider (OpenAI-compatible API)
│   │   ├── noop.py                 # No-op provider (no key configured)
│   │   └── factory.py              # Resolves active provider from config
│   ├── mcp/
│   │   ├── __init__.py
│   │   └── server.py               # Full MCP JSON-RPC 2.0 server (stdio + SSE)
│   ├── parsers/
│   │   ├── __init__.py
│   │   ├── code_parser.py          # AST + regex parser (Python,JS,TS,Go,Rust,Java,C…)
│   │   └── doc_parser.py           # Document parser (PDF, DOCX, MD, TXT)
│   ├── search/
│   │   ├── __init__.py
│   │   ├── embedder.py             # Sentence-transformer embeddings + disk cache
│   │   └── retriever.py            # FAISS / cosine-similarity semantic search
│   ├── storage/
│   │   ├── __init__.py
│   │   ├── db.py                   # SQLite layer (files, chunks, embeddings, chat history)
│   │   └── snapshot.py             # Snapshot create / restore / delete
│   ├── agents/
│   │   ├── __init__.py
│   │   └── context_builder.py      # Context packs for debug/docs/refactor/review/test
│   └── utils/
│       ├── __init__.py
│       ├── display.py              # Rich terminal output helpers
│       └── tokens.py               # tiktoken token counter + cost estimator
├── tests/
│   ├── __init__.py
│   └── test_ragebot.py             # Full pytest test suite (50+ tests)
├── .env.example                    # Example environment variables
├── .gitignore
├── deploy.md                       # Deployment guide (Claude Desktop, Docker, cloud…)
├── pyproject.toml                  # Modern Python packaging (PEP 517)
├── requirements.txt                # Core dependencies
├── setup.py                        # Backward-compatible setup entry
└── README.md                       # This file

______________________________________________________________________

✨ 功能概览

______________________________________________________________________

🚀 快速开始

1.安装

git clone https://github.com/atharvrahate296/Ragebot-MCP
cd Ragebot-MCP

python -m venv .venv
source .venv/bin/activate        # Windows: .venv\Scripts\activate

pip install -e .

2.身份验证

# Gemini — get key at https://aistudio.google.com/apikey
rage auth login gemini
# Prompted: Enter Gemini API key: ••••••••   (stored in OS keyring, never in a file)

# Grok — get key at https://console.x.ai
rage auth login grok

# Check status
rage auth status

# Switch provider
rage auth switch gemini

3.为你的项目建立索引

cd /your/project
rage init
rage save

4.开始使用它

rage ask "Where is the login logic?"
rage chat
rage explain src/auth.py --symbol login
rage docs src/auth.py --output docs/auth.md
rage test src/auth.py --output tests/test_auth.py
rage diff --staged
rage status

______________________________________________________________________

📋 完整的命令参考

核心命令

身份验证子命令

配置子命令

快照子命令

历史记录子命令

MCP子命令

______________________________________________________________________

🤖 大语言模型提供商

谷歌双子座

# Get key: https://aistudio.google.com/apikey
rage auth login gemini

# Configure model
rage config set gemini_model gemini-1.5-pro    # default: gemini-1.5-flash

xAI Grok

# Get key: https://console.x.ai
rage auth login grok

# Configure model
rage config set grok_model grok-3              # default: grok-3-mini

注： Grok在 https://api.x.ai/v1 --除此之外不需要额外的SDK openai.

无LLM（仅用于上下文检索）

rage auth switch none
# rage ask still retrieves and shows context, but won't generate AI answers

______________________________________________________________________

🔌 MCP服务器

RageBot是一个 真实MCP服务器 10个工具：

快速启动（stdio——适用于克劳德桌面/光标）

rage mcp start --transport stdio --project /path/to/project

远程服务器（SSE）

rage mcp start --transport sse --host 0.0.0.0 --port 8765 --project /path/to/project

Claude桌面配置

编辑 ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "ragebot": {
      "command": "/path/to/.venv/bin/python",
      "args": [
        "-m", "ragebot.mcp.server",
        "--project", "/path/to/your/project",
        "--transport", "stdio"
      ]
    }
  }
}

完整部署指南： 看见 deploy.md --涵盖了Claude Desktop、Cursor、Zed、Continue.dev、Docker、Railway、Fly.io、systemd和Nginx/TLS。

______________________________________________________________________

⚙️ 配置参考

设置在 ~/.config/ragebot/config.json.API密钥为 从不 存储在那里。

环境变量

______________________________________________________________________

💬 聊天会话命令

一旦进入 rage chat，使用以下斜线命令：

______________________________________________________________________

🔒 安全

• API密钥使用 OS钥匙圈 --macOS钥匙串、GNOME钥匙圈或Windows凭据管理器
• 钥匙是 从不 写信给 config.json, .env 文件或日志
• rage config show 显示隐藏的按键 ********abcd
• rage config set 将拒绝设置密钥并将您重定向到 rage auth login
• 如果密钥环不可用，则仅通过环境变量设置密钥(GEMINI_API_KEY)

______________________________________________________________________

🛠️ 发展

# Install dev dependencies
pip install -e ".[dev,full]"

# Run tests
pytest tests/ -v

# With coverage report
pytest tests/ -v --cov=ragebot --cov-report=html

# Lint
ruff check ragebot/
black ragebot/

# Type check
mypy ragebot/

______________________________________________________________________

🏗️ 架构概述

rage CLI
   └─► RageBotEngine
          ├── DirectoryScanner     recursive scan, .gitignore logic, file classifier
          ├── CodeParser           Python AST + regex → functions, classes, imports
          ├── DocumentParser       PDF/DOCX/MD/TXT → clean text chunks
          ├── Embedder             sentence-transformers → float vectors (disk cached)
          ├── ContextRetriever     FAISS / cosine similarity → top-k relevant chunks
          ├── Database (SQLite)    files · chunks · embeddings · chat history
          ├── SnapshotManager      versioned DB copies
          ├── LLM Provider         Gemini or Grok (resolved by factory)
          └── ContextBuilder       agent-specific context packs

ragebot.mcp.server
   ├── stdio transport   JSON-RPC 2.0 over stdin/stdout (for desktop clients)
   ├── SSE transport     FastAPI + uvicorn HTTP (for remote / cloud clients)
   └── 10 MCP tools      ask · search · save · explain · file_tree · status
                         export · generate_docs · generate_tests · diff_explain

______________________________________________________________________

📄 许可证

麻省 理工© RageBot 团队

______________________________________________________________________

🔗 链接