AITree MCP Server
Enables AI assistants to generate annotated file trees, retrieve file statistics, list git-changed files, and read file contents from local directories or GitHub repositories.
README
AITree
Project file map generator for AI-assisted development.
Scans a directory and prints a clean, annotated tree that you can paste directly into any AI chat. Instead of uploading your entire codebase, share the map first — then only send the files the AI actually asks for.
Live Demo: paracci.github.io/aitree
webapp/
├── src/
│ ├── components/
│ │ ├── Header.jsx [4KB]
│ │ └── Footer.jsx [2KB]
│ ├── App.jsx [8KB]
│ └── index.js [1KB]
├── public/
│ └── index.html [3KB]
└── package.json [1KB]
# 2 directories, 6 files (18KB total)
#
# File types:
# .jsx 3 ███
# .js 1 █
# .html 1 █
# .json 1 █
Installation
Via pip (recommended):
pip install paracci-aitree
One-liner (Linux / macOS):
curl -fsSL https://raw.githubusercontent.com/paracci/aitree/main/install.sh | bash
One-liner (Windows PowerShell):
irm https://raw.githubusercontent.com/paracci/aitree/main/install.ps1 | iex
From source:
git clone https://github.com/paracci/aitree.git
cd aitree
pip install .
Optional dependencies — install only what you need:
pip install "paracci-aitree[live]" # --live mode (watchdog)
pip install "paracci-aitree[tokens]" # --tokens (tiktoken)
pip install "paracci-aitree[git]" # --git-changed (gitpython)
pip install "paracci-aitree[mcp]" # MCP server (mcp)
pip install "paracci-aitree[all]" # everything + clipboard (pyperclip)
Usage
aitree [path] [options]
path defaults to the current directory if omitted.
Output modes
| Flag | Description |
|---|---|
| (none) | Print tree to stdout |
--save |
Save to _aitree.txt (or _aitree.json / _aitree.md) |
--copy |
Copy output to clipboard |
--live |
Watch for changes and auto-update the saved file |
--serve |
Launch the web UI in your browser |
Output formats
| Flag | Description |
|---|---|
--format text |
Plain text with # comments (default) |
--format json |
Structured JSON with tree + stats + meta |
--format markdown |
Markdown with fenced code block + stats table |
Filtering
| Flag | Description |
|---|---|
--depth N |
Limit tree depth (e.g. --depth 2) |
--include GLOB |
Show only matching files (repeatable) |
--exclude GLOB |
Hide matching files (repeatable) |
--no-gitignore |
Skip .gitignore / .aitreeignore parsing |
--git-changed |
Show only git-changed / untracked files |
Extras
| Flag | Description |
|---|---|
--tokens |
Append estimated token count to output |
--version |
Show version and exit |
Examples
aitree . # current directory → stdout
aitree ~/projects/webapp # specific path → stdout
aitree . --save # save to _aitree.txt
aitree . --copy # copy to clipboard
aitree . --depth 2 # top 2 levels only
aitree . --include "*.py" # Python files only
aitree . --include "*.py" --include "*.js" # multiple patterns
aitree . --exclude "*.test.*" # hide test files
aitree . --git-changed # only modified/untracked files
aitree . --format json --save # save as _aitree.json
aitree . --format markdown --save # save as _aitree.md
aitree . --tokens # append token count
aitree . --live # watch for changes
aitree . --live --depth 2 --git-changed # live + depth + git filter
aitree . --serve # open web UI
aitree . --serve --port 8080 # web UI on a custom port
AI workflow
1. aitree . --copy
2. Paste into your AI chat
3. Ask "Which files do you need?"
4. Send only those files
Web UI
--serve launches a local web server and opens a professional, interactive explorer in your browser.
- Real-time Search: Instant tree filtering with match highlighting.
- Syntax Highlighting: Auto-detection for 30+ languages (via highlight.js).
- Themes: Switch between Dark (Tokyo Night) and Light (GitHub) modes.
- File Preview: Click any file to see content with line numbers and word-wrap.
- Interactive Tree: Fast collapse/expand, depth control, and file-type icons.
- Live Reload: UI auto-refreshes when your project files change.
- Responsive Layout: Resizable sidebar and mobile-friendly design.
aitree . --serve
aitree . --serve --port 8080
MCP Server
AITree ships a Model Context Protocol server so AI assistants (Claude Desktop, etc.) can call it as a tool.
# Install MCP dependency
pip install "paracci-aitree[mcp]"
# Start the server
aitree-mcp
Available tools
| Tool | Description |
|---|---|
aitree_get_tree |
Generate a file tree for a local path or GitHub URL |
aitree_get_stats |
Return structured JSON statistics for a path |
aitree_get_changed |
List git-changed / untracked files |
aitree_read_file |
Read the contents of a file inside the project |
aitree_read_file lets an AI assistant fetch any file it spotted in the tree — without leaving the conversation:
aitree_get_tree(path=".", format="text")
aitree_read_file(path=".", file="src/auth/login.py")
Claude Desktop config (claude_desktop_config.json)
{
"mcpServers": {
"aitree": {
"command": "aitree-mcp"
}
}
}
GitHub URL support
Pass a GitHub repository URL instead of a local path:
aitree https://github.com/owner/repo
aitree https://github.com/owner/repo --depth 2
aitree https://github.com/owner/private-repo --token ghp_xxxx
Works entirely via the GitHub API — no cloning required.
Custom ignore rules
Create .aitreeignore in your project root — same syntax as .gitignore:
# .aitreeignore
*.secret
config/local.json
private/
AITree reads .aitreeignore first, then .gitignore. Use --no-gitignore to skip both.
What gets ignored by default
| Category | Examples |
|---|---|
| VCS & editors | .git, .idea, .vscode, .svn |
| Dependencies | node_modules, .venv, venv, env |
| Build output | dist, build, target, out, bin |
| Cache | __pycache__, .pytest_cache, .mypy_cache |
| Binary & temp | .pyc, .class, .exe, .log, .tmp |
| OS files | .DS_Store, Thumbs.db |
Live mode
--live watches the directory with watchdog and regenerates the output file whenever the structure actually changes. False positives (editor auto-saves, antivirus touches, etc.) are filtered out via content hashing and a 1-second debounce.
🟢 AITree LIVE — Watching: /home/user/projects/webapp [depth=2, git-changed]
Updates trigger only on real changes.
Press Ctrl+C to stop.
[14:32:10] Updated — initial snapshot
[14:33:05] Updated — filesystem change (3 false events skipped)
Combines with any filter flag: --depth, --include, --exclude, --git-changed.
Platform support
| Platform | Tested |
|---|---|
| Linux | ✅ |
| macOS | ✅ |
| Windows | ✅ |
| WSL | ✅ |
Requires Python 3.10+.
License
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。