# Desktop Commander MCP — Local Terminal + Files

> Desktop Commander MCP gives agents controlled local terminal + filesystem tools (including PDF/DOCX/Excel). Install via npx and verify by listing tools.

## Install

Merge the JSON below into your `.mcp.json`:

# Desktop Commander MCP — Local Terminal + Files

> Desktop Commander MCP gives agents controlled local terminal + filesystem tools (including PDF/DOCX/Excel). Install via npx and verify by listing tools.

## Quick Use

1) Install

```bash
npx @wonderwhy-er/desktop-commander@latest setup
```

2) Run

```bash
In your MCP client: ask to list files in a folder and run a safe read-only command
```

3) Verify

```bash
In your MCP client: confirm Desktop Commander tools appear
```

---

## Intro

**Best for**: developers who want a single MCP server for terminal + file inspection while keeping work local

**Works with**: Claude Desktop or any MCP client, Node.js + npx; local filesystem access

**Setup time**: 8 minutes

### Quant Data

- Install via `npx ... setup` with auto-updates (repo)
- Built-in PDF/DOCX/Excel support (repo)

---

## How to Use It Well

Use Desktop Commander as your agent's local bridge: inspect files, run reproducible commands, and keep artifacts (logs, reports) versioned in your repo.

### Adoption Checklist

- Start with one real task and keep the scope narrow
- Capture a baseline: time-to-first-success and output quality
- Version your config/skills so teammates stay in sync

### Guardrails

Define a safety policy: which commands are allowed, which folders are readable, and when human approval is required.

### FAQ

**Q: Is it safe by default?**
A: Treat it like giving a coworker terminal access: start with read-only tasks, restrict directories, and avoid storing secrets in logs.

**Q: Do I need Docker?**
A: No. The repo offers multiple install options; Docker is optional if you want extra isolation.

**Q: How do I avoid context overflow?**
A: Prefer targeted reads and pagination; the repo supports offset/length style reads to keep outputs small.

---

## Source & Thanks

> GitHub: https://github.com/wonderwhy-er/DesktopCommanderMCP
> Owner avatar: https://avatars.githubusercontent.com/u/1150639?v=4
> License (SPDX): MIT
> GitHub stars (verified via `api.github.com/repos/wonderwhy-er/DesktopCommanderMCP`): 6,024


---

<!-- ZH -->

# Desktop Commander MCP——本地终端与文件工具

> Desktop Commander MCP 让 AI 代理在本机安全执行终端与文件操作（含 PDF/DOCX/Excel）。npx 安装后即可在 Claude Desktop/其他 MCP 客户端调用。

## 快速使用

1）安装

```bash
npx @wonderwhy-er/desktop-commander@latest setup
```

2）运行

```bash
In your MCP client: ask to list files in a folder and run a safe read-only command
```

3）验证

```bash
In your MCP client: confirm Desktop Commander tools appear
```

---

## 简介

**适合谁**：想把终端与文件检查统一交给一个 MCP server，并把数据留在本机的开发者

**适用环境**：Claude Desktop 或其他 MCP 客户端、Node.js + npx、本地文件系统权限

**安装耗时**：8 分钟

### 量化信息

- 可用 `npx ... setup` 安装并自动更新（仓库）
- 内置 PDF/DOCX/Excel 支持（仓库）

---

## 用好它的方式

把 Desktop Commander 当作 agent 的本地桥梁：检查文件、跑可复现命令，并把日志/报告作为产物落到仓库里。

### 推广清单

- 先选一个真实任务，小范围试跑
- 记录基线：首次成功耗时与输出质量
- 配置/技能要版本化，避免团队漂移

### 风险与护栏

先定义安全策略：允许哪些命令、可读哪些目录、哪些操作必须人工确认。

### FAQ

**Q: 默认安全吗？**
A: 把它当成给同事终端权限：先做只读任务、限制目录范围，并避免把密钥写进日志。

**Q: 必须用 Docker 吗？**
A: 不必。仓库提供多种安装方式；Docker 只是想要额外隔离时的可选项。

**Q: 怎么避免上下文溢出？**
A: 优先做定点读取和分页；仓库支持按 offset/length 取输出，避免一次塞太多内容。

---

## 来源与感谢

> GitHub：https://github.com/wonderwhy-er/DesktopCommanderMCP
> Owner avatar：https://avatars.githubusercontent.com/u/1150639?v=4
> 许可证（SPDX）：MIT
> GitHub stars（已通过 `api.github.com/repos/wonderwhy-er/DesktopCommanderMCP` 核验）：6,024


---
Source: https://tokrepo.com/en/workflows/desktop-commander-mcp-local-terminal-files
Author: MCP Hub