DevToolkit MCP Server
Provides AI assistants with 28 developer tools across file, git, code analysis, HTTP, and system domains, enabling tasks like file editing, repository management, code analysis, and shell command execution.
README
DevToolkit MCP Server
A production-quality Model Context Protocol (MCP) server that gives AI assistants like Claude 22 powerful developer tools across 5 domains.
Built with TypeScript, strict types, Zod validation, and clean modular architecture.
Features
📁 File Tools (6)
| Tool | Description |
|---|---|
read_file |
Read file contents with optional line range |
write_file |
Write/append to files, auto-creates directories |
list_directory |
List files with metadata (size, modified date) |
search_in_files |
Regex search across files with glob patterns |
get_file_info |
File metadata: size, timestamps, permissions |
delete_file |
Delete files or directories (recursive optional) |
🔧 Git Tools (6)
| Tool | Description |
|---|---|
git_status |
Working tree status, branch info, ahead/behind |
git_log |
Commit history with author, date, message |
git_diff |
Diff between commits, branches, or working tree |
git_blame |
Line-by-line authorship for any file |
git_branches |
List all local and remote branches |
git_show_commit |
Full diff and metadata for a specific commit |
🔍 Code Analysis Tools (6)
| Tool | Description |
|---|---|
analyze_complexity |
Cyclomatic complexity, nesting depth, function count |
find_todos |
Find TODO/FIXME/HACK comments across a codebase |
count_lines |
Code vs comment vs blank line breakdown |
detect_language |
Identify programming language from file extension/shebang |
find_duplicates |
Detect duplicate code blocks across files |
get_imports |
Extract and categorize all import statements |
🌐 HTTP Tools (4)
| Tool | Description |
|---|---|
http_request |
Full HTTP client: GET/POST/PUT/PATCH/DELETE with headers/body |
fetch_json |
Fetch and parse JSON with optional dot-path extraction |
check_url_status |
Batch URL health check with response times |
download_file |
Download files from URLs to local paths |
💻 System Tools (6)
| Tool | Description |
|---|---|
get_system_info |
OS, CPU, memory, Node version, uptime |
get_env |
Read environment variables (sensitive values auto-masked) |
run_command |
Execute shell commands (allowlisted for safety) |
list_processes |
Running processes with CPU/memory usage |
get_disk_usage |
Disk usage for directories |
get_network_info |
Network interfaces and IP addresses |
Installation
git clone https://github.com/your-username/devtoolkit-mcp
cd devtoolkit-mcp
npm install
npm run build
Usage with Claude Desktop
Add to your Claude Desktop config (claude_desktop_config.json):
{
"mcpServers": {
"devtoolkit": {
"command": "node",
"args": ["/absolute/path/to/devtoolkit-mcp/dist/index.js"]
}
}
}
Usage with Cursor / VS Code
Same pattern — point the MCP server config to dist/index.js.
Development
npm run dev # Run with tsx (no build step)
npm run build # Compile TypeScript
npm run lint # ESLint
Advanced MCP Protocol Features
Beyond tools, this server implements the full 2025-06-18 MCP specification:
📂 Resources
Files are exposed as browsable MCP resources — clients (e.g. Claude Desktop) can list, read, and subscribe to changes.
- Paginated listing with cursor-based pagination (50 files/page)
- Resource templates:
file:///{path}for dynamic file access - Subscriptions: clients receive
notifications/resources/updatedwhen a file changes after a write/delete - List-changed notifications: resource list updates after any file mutation
- Annotations: every resource includes
audience,priority, andlastModified
📝 Logging
Structured log notifications sent to the client for every tool call:
- 8 RFC 5424 severity levels:
debug→emergency - Client can set minimum log level via
logging/setLevel - Logs include tool name, args preview, status, and errors
- Sensitive values never logged
🔍 Completion
Autocomplete for file paths in resource URI templates:
- Triggered when client requests
completion/completeon aref/resource - Returns matching file paths up to 20 suggestions with
hasMoreflag
🗣️ Elicitation
Servers request structured user input before destructive operations:
write_file: if the target file exists, asks user to confirm overwrite (with checkbox UI)delete_file: always asks user to confirm + optionally provide a reason- Falls back gracefully if client doesn't support elicitation (proceeds without prompt)
🤖 Sampling
The server can ask the client's LLM to generate completions:
sampleFromClient()helper exported for use in tools- Specifies model preferences (prefers Claude Sonnet, falls back to any Claude)
- Gracefully no-ops if client doesn't support sampling
Architecture
src/
├── index.ts # MCP server, capabilities, tool routing, elicitation, sampling
├── resources.ts # Resources: list, read, subscribe, completion, pagination
├── tools/
│ ├── files.ts # File system tools
│ ├── git.ts # Git integration (simple-git)
│ ├── code-analysis.ts # Static analysis tools
│ ├── http.ts # HTTP client tools (axios)
│ └── system.ts # OS/system tools
└── utils/
├── errors.ts # Typed error classes, helpers
└── logger.ts # Structured log notifications (RFC 5424)
Key design decisions:
- Zod validation on every tool input — no silent failures
- Typed error classes with error codes for structured debugging
- Security:
run_commandhas an explicit allowlist; env vars auto-mask secrets - Cross-platform: works on Windows, macOS, and Linux
Tech Stack
@modelcontextprotocol/sdk— MCP protocolsimple-git— Git operationsaxios— HTTP clientfast-glob— File pattern matchingzod— Runtime schema validation- TypeScript (strict mode)
License
MIT
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。